{ "benchmark": "GeneProv", "version": "v1.1", "frozen": { "v0": "2026-10-10", "v1": "2026-10-11", "v1.1": "2026-10-17" }, "generated": "2026-10-25", "universe": 19256, "strata": { "0-10": 7120, "all": 19256, "zero": 1262, "1-10": 5858, "11-100": 8421, "101-1000": 3301, ">1000": 414 }, "labels_T2": [ "depmap_essential", "hpa_loc_nucleoplasm", "hpa_loc_cytosol", "hpa_loc_plasma_membrane", "hpa_loc_vesicles", "hpa_loc_mitochondria", "hpa_loc_golgi_apparatus", "gtex_tissue_specific", "hpa_tissue_enriched", "clingen_hi", "constraint" ], "index_definition": "mean over T1, T2, T4 of 100*(2*AUROC-1); T3 beside for eligible rows; index4 = mean over the four for eligible rows", "T3_reference": { "linear": { "T3_0-5": 0.7324, "T3_all": 0.7996, "name": "literature-only 2023 text, MPNet" }, "mlp": { "T3_0-5": 0.8774, "T3_all": 0.9029, "name": "literature-only 2023 text, MPNet" } }, "gates": { "R-gate": "PASS", "I-gate1": "PASS", "I-gate2": "PASS", "I-gate3": "PASS" }, "robustness": { "mean_rank": 0.8465, "T2_T4_only": 0.4271, "mlp": 0.9564, "all_genes": 0.8707 }, "weights_spearman": { "0": 0.4436, "1/4": 0.9805, "1/3": 1.0, "1/2": 0.9353 }, "n_families": 3018, "facts": { "universe": 19256, "generated_summaries": 6172, "generated_share": 0.320523, "generated_share_0-10": 0.574438, "generated_by_bin": { "zero": { "n": 1262, "generated": 702 }, "1-10": { "n": 5858, "generated": 3388 }, "11-100": { "n": 8421, "generated": 1864 }, "101-1000": { "n": 3301, "generated": 212 }, ">1000": { "n": 414, "generated": 6 } }, "removal_effect_1-10": { "linear": { "delta": 0.0777, "ci": [ 0.0717, 0.0839 ], "ci_family": [ 0.0705, 0.0857 ], "n": 4310 }, "mlp": { "delta": 0.0334, "ci": [ 0.0302, 0.0364 ], "ci_family": [ 0.0298, 0.0372 ], "n": 4310 } }, "snapshot": "NCBI Gene, 6 October 2026", "n_rows_O_v1": 20, "n_rows_O_v11": 6, "n_rows_F": 6 }, "rows": [ { "track": "O", "tag": "protein16b", "name": "GB.Protein-16B", "short": "GB.Protein-16B", "version": "v1", "in_v0": true, "modality": "protein", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 19133, "meta": { "training": "1.2 trillion amino acids from UniRef90 and ColabFoldDB; masked language modelling", "date": "released Nov 2024; database releases not stated", "signal": "none (sequences only)", "t3": "yes", "t3_why": "sequence rule: no functional annotation or text among its inputs", "source": "https://huggingface.co/genbio-ai/AIDO.Protein-16B", "cite": "Sun et al., 2024" }, "probes": { "linear": { "0-10": { "index": 45.26, "ci": [ 44.51, 46.04 ], "rank": 4, "rank_ci": [ 4, 4 ], "index4": 47.96, "T1": 0.8753, "T2": 0.6763, "T2_labels": 10, "T4": 0.6274, "T1_n": 4878, "T4_n": 6077, "T3": 0.7804, "T3_n": 137, "T3_novel": 0.7429, "skills": { "T1": 75.06, "T2": 35.25, "T4": 25.47, "T3": 56.07 }, "T1_bins": { "zero": 0.9508, "1-10": 0.8654, "11-100": 0.8588, "101-1000": 0.8351, ">1000": 0.796 }, "T2_per_label": { "depmap_essential": 0.811, "hpa_loc_nucleoplasm": 0.6563, "hpa_loc_cytosol": 0.5792, "hpa_loc_plasma_membrane": 0.5682, "hpa_loc_vesicles": 0.5374, "hpa_loc_mitochondria": 0.6974, "hpa_loc_golgi_apparatus": 0.5578, "gtex_tissue_specific": 0.7916, "hpa_tissue_enriched": 0.784, "constraint": 0.7797 }, "T2_per_label_source": "round-14 regeneration", "ci_family": [ 40.05, 49.9 ], "rank_ci_family": [ 3, 4 ], "index_r14": 45.26, "rank_r14": 4, "T4_recall10": 0.0173 }, "all": { "index": 46.19, "ci": [ 45.76, 46.63 ], "rank": 7, "rank_ci": [ 6, 8 ], "index4": 51.29, "T1": 0.8574, "T2": 0.7192, "T2_labels": 11, "T4": 0.6163, "T1_n": null, "T4_n": null, "T3": 0.8329, "T3_n": 347, "T3_novel": 0.7652, "skills": { "T1": 71.48, "T2": 43.83, "T4": 23.25, "T3": 66.59 }, "T2_per_label": { "depmap_essential": 0.8461, "hpa_loc_nucleoplasm": 0.7236, "hpa_loc_cytosol": 0.6231, "hpa_loc_plasma_membrane": 0.6915, "hpa_loc_vesicles": 0.6071, "hpa_loc_mitochondria": 0.7691, "hpa_loc_golgi_apparatus": 0.6166, "gtex_tissue_specific": 0.8416, "hpa_tissue_enriched": 0.8166, "clingen_hi": 0.5577, "constraint": 0.8178 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0165 } }, "mlp": { "0-10": { "index": 54.32, "ci": [ 53.67, 54.94 ], "rank": 2, "rank_ci": [ 2, 3 ], "index4": 61.61, "T1": 0.9446, "T2": 0.7428, "T2_labels": 10, "T4": 0.6274, "T1_n": 4878, "T4_n": 6077, "T3": 0.9174, "T3_n": 137, "T3_novel": 0.8949, "skills": { "T1": 88.92, "T2": 48.56, "T4": 25.47, "T3": 83.49 }, "T1_bins": { "zero": 0.9786, "1-10": 0.9401, "11-100": 0.9261, "101-1000": 0.9034, ">1000": 0.8735 }, "T2_per_label": { "depmap_essential": 0.9422, "hpa_loc_nucleoplasm": 0.6363, "hpa_loc_cytosol": 0.6342, "hpa_loc_plasma_membrane": 0.6665, "hpa_loc_vesicles": 0.6248, "hpa_loc_mitochondria": 0.7579, "hpa_loc_golgi_apparatus": 0.6457, "gtex_tissue_specific": 0.8418, "hpa_tissue_enriched": 0.8117, "constraint": 0.8668 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0173 }, "all": { "index": 53.64, "ci": [ 53.25, 54.06 ], "rank": 5, "rank_ci": [ 4, 6 ], "index4": 61.63, "T1": 0.9258, "T2": 0.7626, "T2_labels": 11, "T4": 0.6163, "T1_n": null, "T4_n": null, "T3": 0.928, "T3_n": 347, "T3_novel": 0.8936, "skills": { "T1": 85.15, "T2": 52.52, "T4": 23.25, "T3": 85.6 }, "T2_per_label": { "depmap_essential": 0.9407, "hpa_loc_nucleoplasm": 0.7265, "hpa_loc_cytosol": 0.648, "hpa_loc_plasma_membrane": 0.7457, "hpa_loc_vesicles": 0.6603, "hpa_loc_mitochondria": 0.8152, "hpa_loc_golgi_apparatus": 0.6689, "gtex_tissue_specific": 0.8732, "hpa_tissue_enriched": 0.8363, "clingen_hi": 0.6033, "constraint": 0.8704 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0165 } } }, "coverage": { "n_present": 19133, "share_missing": 0.0064, "missing_by_bin": { "zero": 16, "1-10": 59, "11-100": 42, "101-1000": 6, ">1000": 0 } } }, { "track": "O", "tag": "esm2_650m", "name": "ESM2-650M", "short": "ESM2-650M", "version": "v1", "in_v0": true, "modality": "protein", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 19133, "meta": { "training": "UniRef50 clusters sampled from UniRef90 (UniRef release 2021_04), masked language modelling", "date": "2021", "signal": "none", "t3": "yes", "t3_why": "sequence rule", "source": "https://huggingface.co/facebook/esm2_t33_650M_UR50D", "cite": "Lin et al., 2023" }, "probes": { "linear": { "0-10": { "index": 44.42, "ci": [ 43.66, 45.12 ], "rank": 5, "rank_ci": [ 5, 6 ], "index4": 47.17, "T1": 0.8792, "T2": 0.6905, "T2_labels": 10, "T4": 0.5967, "T1_n": 4878, "T4_n": 6077, "T3": 0.7771, "T3_n": 137, "T3_novel": 0.7424, "skills": { "T1": 75.83, "T2": 38.1, "T4": 19.33, "T3": 55.42 }, "T1_bins": { "zero": 0.9595, "1-10": 0.8686, "11-100": 0.8604, "101-1000": 0.8381, ">1000": 0.8053 }, "T2_per_label": { "depmap_essential": 0.8412, "hpa_loc_nucleoplasm": 0.6728, "hpa_loc_cytosol": 0.5826, "hpa_loc_plasma_membrane": 0.6136, "hpa_loc_vesicles": 0.5814, "hpa_loc_mitochondria": 0.7103, "hpa_loc_golgi_apparatus": 0.5658, "gtex_tissue_specific": 0.7844, "hpa_tissue_enriched": 0.7858, "constraint": 0.7669 }, "T2_per_label_source": "round-14 regeneration", "ci_family": [ 38.93, 47.97 ], "rank_ci_family": [ 5, 8 ], "index_r14": 43.65, "rank_r14": 6, "T4_recall10": 0.015 }, "all": { "index": 44.94, "ci": [ 44.48, 45.37 ], "rank": 9, "rank_ci": [ 9, 10 ], "index4": 50.66, "T1": 0.8601, "T2": 0.7288, "T2_labels": 11, "T4": 0.5851, "T1_n": null, "T4_n": null, "T3": 0.8391, "T3_n": 347, "T3_novel": 0.7698, "skills": { "T1": 72.03, "T2": 45.75, "T4": 17.03, "T3": 67.82 }, "T2_per_label": { "depmap_essential": 0.864, "hpa_loc_nucleoplasm": 0.7411, "hpa_loc_cytosol": 0.6426, "hpa_loc_plasma_membrane": 0.718, "hpa_loc_vesicles": 0.6467, "hpa_loc_mitochondria": 0.7825, "hpa_loc_golgi_apparatus": 0.6229, "gtex_tissue_specific": 0.8264, "hpa_tissue_enriched": 0.8083, "clingen_hi": 0.5572, "constraint": 0.794 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0129 } }, "mlp": { "0-10": { "index": 52.15, "ci": [ 51.51, 52.79 ], "rank": 5, "rank_ci": [ 5, 7 ], "index4": 59.82, "T1": 0.9465, "T2": 0.7391, "T2_labels": 10, "T4": 0.5967, "T1_n": 4878, "T4_n": 6077, "T3": 0.9142, "T3_n": 137, "T3_novel": 0.8947, "skills": { "T1": 89.3, "T2": 47.82, "T4": 19.33, "T3": 82.85 }, "T1_bins": { "zero": 0.9816, "1-10": 0.9419, "11-100": 0.9291, "101-1000": 0.9092, ">1000": 0.8779 }, "T2_per_label": { "depmap_essential": 0.924, "hpa_loc_nucleoplasm": 0.6642, "hpa_loc_cytosol": 0.6028, "hpa_loc_plasma_membrane": 0.6628, "hpa_loc_vesicles": 0.6305, "hpa_loc_mitochondria": 0.7611, "hpa_loc_golgi_apparatus": 0.6451, "gtex_tissue_specific": 0.8014, "hpa_tissue_enriched": 0.8156, "constraint": 0.8328 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.015 }, "all": { "index": 51.39, "ci": [ 50.99, 51.76 ], "rank": 11, "rank_ci": [ 10, 11 ], "index4": 60.05, "T1": 0.929, "T2": 0.7567, "T2_labels": 11, "T4": 0.5851, "T1_n": null, "T4_n": null, "T3": 0.9301, "T3_n": 347, "T3_novel": 0.8972, "skills": { "T1": 85.8, "T2": 51.33, "T4": 17.03, "T3": 86.02 }, "T2_per_label": { "depmap_essential": 0.9183, "hpa_loc_nucleoplasm": 0.7293, "hpa_loc_cytosol": 0.633, "hpa_loc_plasma_membrane": 0.7368, "hpa_loc_vesicles": 0.6601, "hpa_loc_mitochondria": 0.8053, "hpa_loc_golgi_apparatus": 0.6688, "gtex_tissue_specific": 0.8381, "hpa_tissue_enriched": 0.8154, "clingen_hi": 0.5715, "constraint": 0.8224 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0129 } } }, "coverage": { "n_present": 19133, "share_missing": 0.0064, "missing_by_bin": { "zero": 16, "1-10": 59, "11-100": 42, "101-1000": 6, ">1000": 0 } } }, { "track": "O", "tag": "cell_base", "name": "GB.Cell-100M", "short": "GB.Cell-100M", "version": "v1", "in_v0": true, "modality": "single-cell", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 18593, "meta": { "training": "50 million single-cell transcriptomes from over 100 tissue types, read-depth-aware masked modelling", "date": "released Nov 2024", "signal": "none (expression counts)", "t3": "yes", "t3_why": "expression rule", "source": "https://huggingface.co/genbio-ai/AIDO.Cell-100M", "cite": "Ho et al., 2024" }, "probes": { "linear": { "0-10": { "index": 34.7, "ci": [ 33.91, 35.47 ], "rank": 17, "rank_ci": [ 16, 18 ], "index4": 34.98, "T1": 0.7079, "T2": 0.7327, "T2_labels": 10, "T4": 0.5798, "T1_n": 4878, "T4_n": 6077, "T3": 0.6791, "T3_n": 137, "T3_novel": null, "skills": { "T1": 41.59, "T2": 46.54, "T4": 15.96, "T3": 35.82 }, "T1_bins": { "zero": 0.7733, "1-10": 0.6993, "11-100": 0.7341, "101-1000": 0.7483, ">1000": 0.7396 }, "ci_family": [ 32.81, 36.97 ], "rank_ci_family": [ 15, 18 ], "index_r14": 34.7, "rank_r14": 17, "T4_recall10": 0.0052 }, "all": { "index": 37.62, "ci": [ 37.18, 38.05 ], "rank": 18, "rank_ci": [ 17, 18 ], "index4": 39.14, "T1": 0.7294, "T2": 0.7567, "T2_labels": 11, "T4": 0.5781, "T1_n": null, "T4_n": null, "T3": 0.7185, "T3_n": 347, "T3_novel": null, "skills": { "T1": 45.88, "T2": 51.35, "T4": 15.62, "T3": 43.7 }, "T4_recall10": 0.0078 } }, "mlp": { "0-10": { "index": 44.85, "ci": [ 44.24, 45.5 ], "rank": 16, "rank_ci": [ 16, 17 ], "index4": 51.5, "T1": 0.8659, "T2": 0.727, "T2_labels": 10, "T4": 0.5798, "T1_n": 4878, "T4_n": 6077, "T3": 0.8573, "T3_n": 137, "T3_novel": null, "skills": { "T1": 73.18, "T2": 45.41, "T4": 15.96, "T3": 71.46 }, "T1_bins": { "zero": 0.9204, "1-10": 0.8587, "11-100": 0.8528, "101-1000": 0.8422, ">1000": 0.8188 }, "T4_recall10": 0.0052 }, "all": { "index": 45.61, "ci": [ 45.21, 45.99 ], "rank": 17, "rank_ci": [ 16, 18 ], "index4": 51.6, "T1": 0.8537, "T2": 0.7523, "T2_labels": 11, "T4": 0.5781, "T1_n": null, "T4_n": null, "T3": 0.8479, "T3_n": 347, "T3_novel": null, "skills": { "T1": 70.74, "T2": 50.46, "T4": 15.62, "T3": 69.59 }, "T4_recall10": 0.0078 } } }, "coverage": { "n_present": 18593, "share_missing": 0.0344, "missing_by_bin": { "zero": 295, "1-10": 277, "11-100": 71, "101-1000": 19, ">1000": 1 } } }, { "track": "O", "tag": "geneformer", "name": "Geneformer V2-316M", "short": "Geneformer V2-316M", "version": "v1", "in_v0": true, "modality": "single-cell", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 19229, "meta": { "training": "about 104 million human single-cell transcriptomes (Genecorpus-104M), rank-value encoding", "date": "trained Dec 2024", "signal": "none", "t3": "yes", "t3_why": "expression rule", "source": "https://huggingface.co/ctheodoris/Geneformer", "cite": "Chen et al., 2024" }, "probes": { "linear": { "0-10": { "index": 36.59, "ci": [ 35.75, 37.34 ], "rank": 14, "rank_ci": [ 14, 15 ], "index4": 35.76, "T1": 0.7275, "T2": 0.7333, "T2_labels": 10, "T4": 0.5879, "T1_n": 4878, "T4_n": 6077, "T3": 0.6664, "T3_n": 137, "T3_novel": null, "skills": { "T1": 45.51, "T2": 46.67, "T4": 17.59, "T3": 33.28 }, "T1_bins": { "zero": 0.819, "1-10": 0.7155, "11-100": 0.7307, "101-1000": 0.7388, ">1000": 0.7106 }, "ci_family": [ 34.31, 39.21 ], "rank_ci_family": [ 13, 16 ], "index_r14": 36.59, "rank_r14": 14, "T4_recall10": 0.0038 }, "all": { "index": 37.73, "ci": [ 37.29, 38.14 ], "rank": 17, "rank_ci": [ 17, 18 ], "index4": 37.79, "T1": 0.7309, "T2": 0.7455, "T2_labels": 11, "T4": 0.5896, "T1_n": null, "T4_n": null, "T3": 0.6898, "T3_n": 347, "T3_novel": null, "skills": { "T1": 46.17, "T2": 49.11, "T4": 17.91, "T3": 37.97 }, "T4_recall10": 0.0042 } }, "mlp": { "0-10": { "index": 45.94, "ci": [ 45.27, 46.58 ], "rank": 14, "rank_ci": [ 14, 15 ], "index4": 52.16, "T1": 0.8636, "T2": 0.7375, "T2_labels": 10, "T4": 0.5879, "T1_n": 4878, "T4_n": 6077, "T3": 0.8541, "T3_n": 137, "T3_novel": null, "skills": { "T1": 72.73, "T2": 47.5, "T4": 17.59, "T3": 70.82 }, "T1_bins": { "zero": 0.9353, "1-10": 0.8542, "11-100": 0.8387, "101-1000": 0.8274, ">1000": 0.7929 }, "T4_recall10": 0.0038 }, "all": { "index": 45.37, "ci": [ 44.98, 45.77 ], "rank": 18, "rank_ci": [ 17, 18 ], "index4": 51.37, "T1": 0.8426, "T2": 0.7484, "T2_labels": 11, "T4": 0.5896, "T1_n": null, "T4_n": null, "T3": 0.8468, "T3_n": 347, "T3_novel": null, "skills": { "T1": 68.52, "T2": 49.69, "T4": 17.91, "T3": 69.37 }, "T4_recall10": 0.0042 } } }, "coverage": { "n_present": 19229, "share_missing": 0.0014, "missing_by_bin": { "zero": 4, "1-10": 11, "11-100": 10, "101-1000": 2, ">1000": 0 } } }, { "track": "O", "tag": "genept_ada", "name": "GenePT ada-002 (NCBI text, as released)", "short": "GenePT ada-002", "version": "v1", "in_v0": true, "modality": "text", "cls": "gen", "text": "genept", "train_text": null, "encoder": "text-embedding-ada-002", "text_label": "31% of the embedded genes carry a GO-generated summary", "gen_share": 0.3089, "n_genes": 18942, "meta": { "training": "NCBI Gene summaries (October 2023 snapshot) embedded with OpenAI text-embedding-ada-002", "date": "text Oct 2023; encoder 2022", "signal": "31% of the embedded summaries are generated from GO", "t3": "yes", "t3_why": "admitted under the tag-date proxy; a pre-2023 tag is not a verified pre-cutoff text snapshot", "source": "https://github.com/yiqunchen/GenePT", "cite": "Chen et al., 2024" }, "probes": { "linear": { "0-10": { "index": 41.44, "ci": [ 40.66, 42.19 ], "rank": 11, "rank_ci": [ 10, 11 ], "index4": 41.09, "T1": 0.8992, "T2": 0.6606, "T2_labels": 10, "T4": 0.5618, "T1_n": 4878, "T4_n": 6077, "T3": 0.7001, "T3_n": 137, "T3_novel": 0.6409, "skills": { "T1": 79.84, "T2": 32.12, "T4": 12.37, "T3": 40.02 }, "T1_bins": { "zero": 0.9435, "1-10": 0.8933, "11-100": 0.8678, "101-1000": 0.847, ">1000": 0.8245 }, "T2_per_label": { "depmap_essential": 0.8214, "hpa_loc_nucleoplasm": 0.6571, "hpa_loc_cytosol": 0.5631, "hpa_loc_plasma_membrane": 0.6266, "hpa_loc_vesicles": 0.5822, "hpa_loc_mitochondria": 0.6968, "hpa_loc_golgi_apparatus": 0.5979, "gtex_tissue_specific": 0.7232, "hpa_tissue_enriched": 0.6971, "constraint": 0.641 }, "T2_per_label_source": "round-14 regeneration", "ci_family": [ 37.89, 44.88 ], "rank_ci_family": [ 9, 12 ], "index_r14": 41.44, "rank_r14": 11, "T4_recall10": 0.0167 }, "all": { "index": 42.54, "ci": [ 42.08, 43.02 ], "rank": 12, "rank_ci": [ 12, 13 ], "index4": 46.11, "T1": 0.8718, "T2": 0.701, "T2_labels": 11, "T4": 0.5653, "T1_n": null, "T4_n": null, "T3": 0.784, "T3_n": 347, "T3_novel": 0.7059, "skills": { "T1": 74.36, "T2": 40.21, "T4": 13.05, "T3": 56.8 }, "T2_per_label": { "depmap_essential": 0.8267, "hpa_loc_nucleoplasm": 0.7109, "hpa_loc_cytosol": 0.6012, "hpa_loc_plasma_membrane": 0.6962, "hpa_loc_vesicles": 0.6197, "hpa_loc_mitochondria": 0.7589, "hpa_loc_golgi_apparatus": 0.6436, "gtex_tissue_specific": 0.792, "hpa_tissue_enriched": 0.7633, "clingen_hi": 0.5995, "constraint": 0.6995 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0217 } }, "mlp": { "0-10": { "index": 47.31, "ci": [ 46.58, 48.0 ], "rank": 12, "rank_ci": [ 11, 13 ], "index4": 55.76, "T1": 0.9536, "T2": 0.6943, "T2_labels": 10, "T4": 0.5618, "T1_n": 4878, "T4_n": 6077, "T3": 0.9054, "T3_n": 137, "T3_novel": 0.8795, "skills": { "T1": 90.71, "T2": 38.86, "T4": 12.37, "T3": 81.09 }, "T1_bins": { "zero": 0.9775, "1-10": 0.9504, "11-100": 0.9272, "101-1000": 0.9019, ">1000": 0.8705 }, "T2_per_label": { "depmap_essential": 0.8893, "hpa_loc_nucleoplasm": 0.6267, "hpa_loc_cytosol": 0.5631, "hpa_loc_plasma_membrane": 0.6774, "hpa_loc_vesicles": 0.6378, "hpa_loc_mitochondria": 0.7201, "hpa_loc_golgi_apparatus": 0.654, "gtex_tissue_specific": 0.7635, "hpa_tissue_enriched": 0.7638, "constraint": 0.6471 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0167 }, "all": { "index": 47.9, "ci": [ 47.49, 48.31 ], "rank": 15, "rank_ci": [ 15, 15 ], "index4": 56.72, "T1": 0.9285, "T2": 0.7246, "T2_labels": 11, "T4": 0.5653, "T1_n": null, "T4_n": null, "T3": 0.9159, "T3_n": 347, "T3_novel": 0.8831, "skills": { "T1": 85.71, "T2": 44.93, "T4": 13.05, "T3": 83.18 }, "T2_per_label": { "depmap_essential": 0.8802, "hpa_loc_nucleoplasm": 0.6965, "hpa_loc_cytosol": 0.6113, "hpa_loc_plasma_membrane": 0.7258, "hpa_loc_vesicles": 0.6535, "hpa_loc_mitochondria": 0.7852, "hpa_loc_golgi_apparatus": 0.6765, "gtex_tissue_specific": 0.8065, "hpa_tissue_enriched": 0.786, "clingen_hi": 0.6372, "constraint": 0.7123 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0217 } } }, "coverage": { "n_present": 18942, "share_missing": 0.0163, "missing_by_bin": { "zero": 85, "1-10": 168, "11-100": 46, "101-1000": 13, ">1000": 2 } } }, { "track": "O", "tag": "genept_3large", "name": "GenePT text-embedding-3-large (as released)", "short": "GenePT 3-large", "version": "v1", "in_v0": true, "modality": "text", "cls": "gen", "text": "genept", "train_text": null, "encoder": "text-embedding-3-large", "text_label": "31% of the embedded genes carry a GO-generated summary", "gen_share": 0.3088, "n_genes": 18945, "meta": { "training": "the same NCBI text plus UniProt summaries, text-embedding-3-large", "date": "text 2024; encoder 2024", "signal": "generated summaries and UniProt text", "t3": "no", "t3_why": "text and encoder after the cutoff", "source": "https://github.com/yiqunchen/GenePT", "cite": "Chen et al., 2024" }, "probes": { "linear": { "0-10": { "index": 43.04, "ci": [ 42.28, 43.76 ], "rank": 7, "rank_ci": [ 7, 9 ], "index4": null, "T1": 0.9187, "T2": 0.6557, "T2_labels": 10, "T4": 0.5713, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.74, "T2": 31.13, "T4": 14.26, "T3": null }, "T1_bins": { "zero": 0.9457, "1-10": 0.9151, "11-100": 0.8985, "101-1000": 0.8711, ">1000": 0.8538 }, "ci_family": [ 38.6, 46.9 ], "rank_ci_family": [ 5, 9 ], "index_r14": 43.04, "rank_r14": 7, "T4_recall10": 0.0192 }, "all": { "index": 46.01, "ci": [ 45.55, 46.46 ], "rank": 8, "rank_ci": [ 7, 8 ], "index4": null, "T1": 0.8979, "T2": 0.7126, "T2_labels": 11, "T4": 0.5796, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 79.58, "T2": 42.52, "T4": 15.91, "T3": null }, "T4_recall10": 0.0269 } }, "mlp": { "0-10": { "index": 50.87, "ci": [ 50.23, 51.56 ], "rank": 9, "rank_ci": [ 9, 9 ], "index4": null, "T1": 0.9641, "T2": 0.7277, "T2_labels": 10, "T4": 0.5713, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 92.82, "T2": 45.55, "T4": 14.26, "T3": null }, "T1_bins": { "zero": 0.9803, "1-10": 0.9619, "11-100": 0.9428, "101-1000": 0.9159, ">1000": 0.8867 }, "T4_recall10": 0.0192 }, "all": { "index": 52.57, "ci": [ 52.14, 52.98 ], "rank": 7, "rank_ci": [ 7, 8 ], "index4": null, "T1": 0.9423, "T2": 0.7666, "T2_labels": 11, "T4": 0.5796, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 88.47, "T2": 53.32, "T4": 15.91, "T3": null }, "T4_recall10": 0.0269 } } }, "coverage": { "n_present": 18945, "share_missing": 0.0162, "missing_by_bin": { "zero": 85, "1-10": 166, "11-100": 45, "101-1000": 13, ">1000": 2 } } }, { "track": "O", "tag": "scelmo_gpt35", "name": "scELMo (GPT-3.5-written descriptions)", "short": "scELMo (GPT-3.5 text)", "version": "v1", "in_v0": true, "modality": "text", "cls": "clean", "text": "llm", "train_text": null, "encoder": "text-embedding-ada-002", "text_label": "LLM-written descriptions; no NCBI or UniProt text", "gen_share": 0.0, "n_genes": 19231, "meta": { "training": "descriptions written by GPT-3.5 from its own memory, embedded with ada-002", "date": "GPT-3.5 training data to 2021", "signal": "LLM memory of the literature and databases", "t3": "yes", "t3_why": "admitted under the reported pre-2022 training-data assumption; not a verified historical corpus", "source": "https://github.com/HelloWorldLTY/scELMo", "cite": "Liu et al., 2023" }, "probes": { "linear": { "0-10": { "index": 36.08, "ci": [ 35.23, 36.95 ], "rank": 15, "rank_ci": [ 14, 16 ], "index4": 38.72, "T1": 0.8393, "T2": 0.6428, "T2_labels": 10, "T4": 0.5591, "T1_n": 4878, "T4_n": 6077, "T3": 0.7332, "T3_n": 137, "T3_novel": null, "skills": { "T1": 67.87, "T2": 28.56, "T4": 11.81, "T3": 46.64 }, "T1_bins": { "zero": 0.9274, "1-10": 0.8277, "11-100": 0.8618, "101-1000": 0.8633, ">1000": 0.8468 }, "ci_family": [ 32.62, 39.65 ], "rank_ci_family": [ 14, 16 ], "index_r14": 36.08, "rank_r14": 15, "T4_recall10": 0.0165 }, "all": { "index": 42.28, "ci": [ 41.79, 42.72 ], "rank": 13, "rank_ci": [ 12, 13 ], "index4": 47.57, "T1": 0.8552, "T2": 0.7072, "T2_labels": 11, "T4": 0.5718, "T1_n": null, "T4_n": null, "T3": 0.8172, "T3_n": 347, "T3_novel": null, "skills": { "T1": 71.03, "T2": 41.45, "T4": 14.36, "T3": 63.44 }, "T4_recall10": 0.0238 } }, "mlp": { "0-10": { "index": 44.76, "ci": [ 44.02, 45.51 ], "rank": 17, "rank_ci": [ 16, 17 ], "index4": 53.56, "T1": 0.9322, "T2": 0.6802, "T2_labels": 10, "T4": 0.5591, "T1_n": 4878, "T4_n": 6077, "T3": 0.8998, "T3_n": 137, "T3_novel": null, "skills": { "T1": 86.44, "T2": 36.04, "T4": 11.81, "T3": 79.96 }, "T1_bins": { "zero": 0.9746, "1-10": 0.9266, "11-100": 0.9292, "101-1000": 0.9142, ">1000": 0.8865 }, "T4_recall10": 0.0165 }, "all": { "index": 48.58, "ci": [ 48.15, 48.98 ], "rank": 13, "rank_ci": [ 12, 14 ], "index4": 57.61, "T1": 0.9261, "T2": 0.7308, "T2_labels": 11, "T4": 0.5718, "T1_n": null, "T4_n": null, "T3": 0.9235, "T3_n": 347, "T3_novel": null, "skills": { "T1": 85.22, "T2": 46.16, "T4": 14.36, "T3": 84.7 }, "T4_recall10": 0.0238 } } }, "coverage": { "n_present": 19231, "share_missing": 0.0013, "missing_by_bin": { "zero": 10, "1-10": 13, "11-100": 2, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "llm_qwen25", "name": "Qwen2.5-7B-written descriptions", "short": "Qwen2.5-written text", "version": "v1", "in_v0": true, "modality": "text", "cls": "clean", "text": "llm", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "LLM-written descriptions; no NCBI or UniProt text", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "descriptions written by Qwen2.5-7B-Instruct, embedded with Qwen3-Embedding-8B", "date": "LLM 2024; encoder 2025", "signal": "LLM memory", "t3": "no", "t3_why": "LLM and encoder after the cutoff", "source": "https://huggingface.co/Qwen/Qwen2.5-7B-Instruct", "cite": "Qwen, 2024" }, "probes": { "linear": { "0-10": { "index": 31.19, "ci": [ 30.35, 32.05 ], "rank": 19, "rank_ci": [ 19, 20 ], "index4": null, "T1": 0.8188, "T2": 0.5997, "T2_labels": 10, "T4": 0.5493, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 63.77, "T2": 19.94, "T4": 9.87, "T3": null }, "T1_bins": { "zero": 0.9366, "1-10": 0.8033, "11-100": 0.8136, "101-1000": 0.8357, ">1000": 0.8412 }, "ci_family": [ 27.52, 34.84 ], "rank_ci_family": [ 19, 20 ], "index_r14": 31.19, "rank_r14": 19, "T4_recall10": 0.0164 }, "all": { "index": 35.74, "ci": [ 35.25, 36.21 ], "rank": 19, "rank_ci": [ 19, 19 ], "index4": null, "T1": 0.8202, "T2": 0.6585, "T2_labels": 11, "T4": 0.5574, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 64.03, "T2": 31.7, "T4": 11.47, "T3": null }, "T4_recall10": 0.0208 } }, "mlp": { "0-10": { "index": 40.36, "ci": [ 39.59, 41.09 ], "rank": 19, "rank_ci": [ 19, 19 ], "index4": null, "T1": 0.9063, "T2": 0.6498, "T2_labels": 10, "T4": 0.5493, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 81.26, "T2": 29.95, "T4": 9.87, "T3": null }, "T1_bins": { "zero": 0.971, "1-10": 0.8978, "11-100": 0.8896, "101-1000": 0.8858, ">1000": 0.8686 }, "T4_recall10": 0.0164 }, "all": { "index": 44.04, "ci": [ 43.59, 44.48 ], "rank": 19, "rank_ci": [ 19, 19 ], "index4": null, "T1": 0.8932, "T2": 0.71, "T2_labels": 11, "T4": 0.5574, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 78.64, "T2": 42.01, "T4": 11.47, "T3": null }, "T4_recall10": 0.0208 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "text_refseq", "name": "Rich summary-filtered text", "short": "Rich summary-filtered text", "version": "v1", "in_v0": true, "modality": "text", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Summary-filtered text (generated summaries removed); includes UniProt function text", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "NCBI RefSeq summaries and UniProt function text of October 2026, embedded with Qwen3-Embedding-8B", "date": "text 2026; encoder 2025", "signal": "UniProt text (co-curated); encoder memory may hold NCBI pages newer than the cutoff", "t3": "no", "t3_why": "text after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 42.81, "ci": [ 42.01, 43.59 ], "rank": 8, "rank_ci": [ 7, 9 ], "index4": null, "T1": 0.9153, "T2": 0.6509, "T2_labels": 10, "T4": 0.5759, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.06, "T2": 30.18, "T4": 15.19, "T3": null }, "T1_bins": { "zero": 0.9411, "1-10": 0.9119, "11-100": 0.9126, "101-1000": 0.8905, ">1000": 0.8653 }, "T2_per_label": { "depmap_essential": 0.819, "hpa_loc_nucleoplasm": 0.6353, "hpa_loc_cytosol": 0.5329, "hpa_loc_plasma_membrane": 0.5812, "hpa_loc_vesicles": 0.5254, "hpa_loc_mitochondria": 0.692, "hpa_loc_golgi_apparatus": 0.5581, "gtex_tissue_specific": 0.7707, "hpa_tissue_enriched": 0.7726, "constraint": 0.622 }, "T2_per_label_source": "round-14 regeneration", "ci_family": [ 38.73, 46.4 ], "rank_ci_family": [ 6, 9 ], "index_r14": 42.81, "rank_r14": 8, "T4_recall10": 0.0203 }, "all": { "index": 47.81, "ci": [ 47.39, 48.26 ], "rank": 4, "rank_ci": [ 3, 4 ], "index4": null, "T1": 0.9079, "T2": 0.7221, "T2_labels": 11, "T4": 0.5872, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 81.58, "T2": 44.42, "T4": 17.43, "T3": null }, "T2_per_label": { "depmap_essential": 0.864, "hpa_loc_nucleoplasm": 0.6998, "hpa_loc_cytosol": 0.6046, "hpa_loc_plasma_membrane": 0.6801, "hpa_loc_vesicles": 0.6089, "hpa_loc_mitochondria": 0.7757, "hpa_loc_golgi_apparatus": 0.6225, "gtex_tissue_specific": 0.8404, "hpa_tissue_enriched": 0.8229, "clingen_hi": 0.7111, "constraint": 0.7133 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0306 } }, "mlp": { "0-10": { "index": 52.14, "ci": [ 51.45, 52.84 ], "rank": 6, "rank_ci": [ 5, 7 ], "index4": null, "T1": 0.9674, "T2": 0.7388, "T2_labels": 10, "T4": 0.5759, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 93.47, "T2": 47.76, "T4": 15.19, "T3": null }, "T1_bins": { "zero": 0.9822, "1-10": 0.9654, "11-100": 0.952, "101-1000": 0.9263, ">1000": 0.8944 }, "T2_per_label": { "depmap_essential": 0.9346, "hpa_loc_nucleoplasm": 0.6668, "hpa_loc_cytosol": 0.6199, "hpa_loc_plasma_membrane": 0.6923, "hpa_loc_vesicles": 0.6459, "hpa_loc_mitochondria": 0.7669, "hpa_loc_golgi_apparatus": 0.7039, "gtex_tissue_specific": 0.8333, "hpa_tissue_enriched": 0.8348, "constraint": 0.6899 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0203 }, "all": { "index": 55.12, "ci": [ 54.72, 55.51 ], "rank": 2, "rank_ci": [ 2, 2 ], "index4": null, "T1": 0.95, "T2": 0.7896, "T2_labels": 11, "T4": 0.5872, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 90.0, "T2": 57.91, "T4": 17.43, "T3": null }, "T2_per_label": { "depmap_essential": 0.9568, "hpa_loc_nucleoplasm": 0.7492, "hpa_loc_cytosol": 0.665, "hpa_loc_plasma_membrane": 0.7755, "hpa_loc_vesicles": 0.7003, "hpa_loc_mitochondria": 0.8431, "hpa_loc_golgi_apparatus": 0.7432, "gtex_tissue_specific": 0.885, "hpa_tissue_enriched": 0.8689, "clingen_hi": 0.7252, "constraint": 0.773 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0306 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "text_refseq_nomenfree", "name": "Rich summary-filtered text, names masked", "short": "Summary-filtered, names masked", "version": "v1", "in_v0": true, "modality": "text", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Summary-filtered text (generated summaries removed); includes UniProt function text", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "NCBI RefSeq summaries and UniProt function text of October 2026, embedded with Qwen3-Embedding-8B; gene names and symbols masked", "date": "text 2026; encoder 2025", "signal": "UniProt text (co-curated)", "t3": "no", "t3_why": "text after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 39.22, "ci": [ 38.36, 40.11 ], "rank": 13, "rank_ci": [ 12, 13 ], "index4": null, "T1": 0.8588, "T2": 0.6549, "T2_labels": 10, "T4": 0.5746, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 71.76, "T2": 30.99, "T4": 14.91, "T3": null }, "T1_bins": { "zero": 0.8136, "1-10": 0.8647, "11-100": 0.9067, "101-1000": 0.8857, ">1000": 0.864 }, "ci_family": [ 36.08, 42.37 ], "rank_ci_family": [ 12, 13 ], "index_r14": 39.22, "rank_r14": 13, "T4_recall10": 0.017 }, "all": { "index": 46.9, "ci": [ 46.47, 47.31 ], "rank": 5, "rank_ci": [ 5, 6 ], "index4": null, "T1": 0.8875, "T2": 0.7292, "T2_labels": 11, "T4": 0.5867, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 77.51, "T2": 45.84, "T4": 17.34, "T3": null }, "T4_recall10": 0.0294 } }, "mlp": { "0-10": { "index": 47.88, "ci": [ 47.11, 48.59 ], "rank": 11, "rank_ci": [ 11, 12 ], "index4": null, "T1": 0.9461, "T2": 0.6975, "T2_labels": 10, "T4": 0.5746, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 89.21, "T2": 39.5, "T4": 14.91, "T3": null }, "T1_bins": { "zero": 0.93, "1-10": 0.9482, "11-100": 0.9502, "101-1000": 0.9261, ">1000": 0.8922 }, "T4_recall10": 0.017 }, "all": { "index": 53.51, "ci": [ 53.1, 53.93 ], "rank": 6, "rank_ci": [ 5, 6 ], "index4": null, "T1": 0.9428, "T2": 0.7732, "T2_labels": 11, "T4": 0.5867, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 88.57, "T2": 54.63, "T4": 17.34, "T3": null }, "T4_recall10": 0.0294 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "t23a__qwen8b", "name": "GenePT text as found", "short": "GenePT text as found", "version": "v1", "in_v0": true, "modality": "text", "cls": "gen", "text": "genept", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "30% of the embedded genes carry a GO-generated summary", "gen_share": 0.3039, "n_genes": 19256, "meta": { "training": "GenePT's October-2023 NCBI text as released, re-encoded with Qwen3-Embedding-8B", "date": "text Oct 2023; encoder 2025", "signal": "generated summaries (as found)", "t3": "no", "t3_why": "encoder after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 41.5, "ci": [ 40.71, 42.21 ], "rank": 10, "rank_ci": [ 10, 11 ], "index4": null, "T1": 0.9044, "T2": 0.6533, "T2_labels": 10, "T4": 0.5647, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 80.89, "T2": 30.66, "T4": 12.95, "T3": null }, "T1_bins": { "zero": 0.9616, "1-10": 0.8969, "11-100": 0.8816, "101-1000": 0.8618, ">1000": 0.8374 }, "ci_family": [ 37.2, 45.27 ], "rank_ci_family": [ 10, 11 ], "index_r14": 41.53, "rank_r14": 10, "T4_recall10": 0.0149 }, "all": { "index": 44.43, "ci": [ 43.96, 44.87 ], "rank": 11, "rank_ci": [ 9, 11 ], "index4": null, "T1": 0.8833, "T2": 0.7125, "T2_labels": 11, "T4": 0.5707, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 76.66, "T2": 42.5, "T4": 14.14, "T3": null }, "T4_recall10": 0.0228 } }, "mlp": { "0-10": { "index": 50.04, "ci": [ 49.35, 50.73 ], "rank": 10, "rank_ci": [ 10, 10 ], "index4": null, "T1": 0.9557, "T2": 0.7302, "T2_labels": 10, "T4": 0.5647, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 91.13, "T2": 46.04, "T4": 12.95, "T3": null }, "T1_bins": { "zero": 0.9856, "1-10": 0.9517, "11-100": 0.9281, "101-1000": 0.9027, ">1000": 0.8714 }, "T4_recall10": 0.0149 }, "all": { "index": 51.77, "ci": [ 51.39, 52.18 ], "rank": 9, "rank_ci": [ 8, 11 ], "index4": null, "T1": 0.9298, "T2": 0.7761, "T2_labels": 11, "T4": 0.5707, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 85.95, "T2": 55.21, "T4": 14.14, "T3": null }, "T4_recall10": 0.0228 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "t23l__qwen8b", "name": "GenePT text, generated summaries removed", "short": "GenePT text, generated removed", "version": "v1", "in_v0": true, "modality": "text", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Summary-filtered text (generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "GenePT's October-2023 NCBI text with generated summaries removed (corpus L), Qwen3-Embedding-8B", "date": "text Oct 2023; encoder 2025", "signal": "curated prose and encoder memory remain; generated summaries removed", "t3": "no", "t3_why": "encoder after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 34.19, "ci": [ 33.28, 35.08 ], "rank": 18, "rank_ci": [ 17, 18 ], "index4": null, "T1": 0.8325, "T2": 0.6242, "T2_labels": 10, "T4": 0.5561, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 66.49, "T2": 24.85, "T4": 11.22, "T3": null }, "T1_bins": { "zero": 0.9349, "1-10": 0.819, "11-100": 0.8551, "101-1000": 0.8591, ">1000": 0.8427 }, "ci_family": [ 30.53, 37.77 ], "rank_ci_family": [ 17, 18 ], "index_r14": 34.16, "rank_r14": 18, "T4_recall10": 0.0135 }, "all": { "index": 40.7, "ci": [ 40.23, 41.19 ], "rank": 14, "rank_ci": [ 14, 15 ], "index4": null, "T1": 0.849, "T2": 0.7085, "T2_labels": 11, "T4": 0.5531, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 69.79, "T2": 41.69, "T4": 10.61, "T3": null }, "T4_recall10": 0.0208 } }, "mlp": { "0-10": { "index": 43.95, "ci": [ 43.22, 44.75 ], "rank": 18, "rank_ci": [ 17, 18 ], "index4": null, "T1": 0.9245, "T2": 0.6786, "T2_labels": 10, "T4": 0.5561, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.91, "T2": 35.72, "T4": 11.22, "T3": null }, "T1_bins": { "zero": 0.9721, "1-10": 0.9183, "11-100": 0.9231, "101-1000": 0.9117, ">1000": 0.8812 }, "T4_recall10": 0.0135 }, "all": { "index": 48.64, "ci": [ 48.22, 49.06 ], "rank": 12, "rank_ci": [ 12, 14 ], "index4": null, "T1": 0.9202, "T2": 0.7562, "T2_labels": 11, "T4": 0.5531, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.05, "T2": 51.25, "T4": 10.61, "T3": null }, "T4_recall10": 0.0208 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "text_symbol", "name": "Gene symbol only", "short": "Gene symbol only", "version": "v1", "in_v0": true, "modality": "text", "cls": "clean", "text": "symbol", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Gene symbol only", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "the HGNC symbol alone, embedded with Qwen3-Embedding-8B", "date": "encoder 2025", "signal": "none in the text; encoder memory", "t3": "no", "t3_why": "encoder after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 35.6, "ci": [ 34.71, 36.4 ], "rank": 16, "rank_ci": [ 15, 17 ], "index4": null, "T1": 0.8539, "T2": 0.6232, "T2_labels": 10, "T4": 0.5569, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 70.78, "T2": 24.64, "T4": 11.38, "T3": null }, "T1_bins": { "zero": 0.9361, "1-10": 0.8431, "11-100": 0.8444, "101-1000": 0.8479, ">1000": 0.8494 }, "ci_family": [ 31.97, 39.15 ], "rank_ci_family": [ 14, 17 ], "index_r14": 35.6, "rank_r14": 16, "T4_recall10": 0.0181 }, "all": { "index": 39.3, "ci": [ 38.82, 39.73 ], "rank": 16, "rank_ci": [ 16, 16 ], "index4": null, "T1": 0.848, "T2": 0.6758, "T2_labels": 11, "T4": 0.5657, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 69.6, "T2": 35.16, "T4": 13.14, "T3": null }, "T4_recall10": 0.0243 } }, "mlp": { "0-10": { "index": 45.86, "ci": [ 45.12, 46.57 ], "rank": 15, "rank_ci": [ 14, 15 ], "index4": null, "T1": 0.933, "T2": 0.698, "T2_labels": 10, "T4": 0.5569, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 86.6, "T2": 39.59, "T4": 11.38, "T3": null }, "T1_bins": { "zero": 0.9806, "1-10": 0.9267, "11-100": 0.9169, "101-1000": 0.9019, ">1000": 0.8813 }, "T4_recall10": 0.0181 }, "all": { "index": 48.55, "ci": [ 48.11, 48.95 ], "rank": 14, "rank_ci": [ 12, 14 ], "index4": null, "T1": 0.9178, "T2": 0.7448, "T2_labels": 11, "T4": 0.5657, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.55, "T2": 48.96, "T4": 13.14, "T3": null }, "T4_recall10": 0.0243 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "dna300m", "name": "GB.DNA-300M", "short": "GB.DNA-300M", "version": "v1", "in_v0": false, "modality": "dna", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 19236, "meta": { "training": "genomes of 796 species (10.6 billion nucleotides; NCBI assemblies)", "date": "2024", "signal": "none", "t3": "yes", "t3_why": "sequence rule", "source": "https://huggingface.co/genbio-ai/AIDO.DNA-300M", "cite": "GenBio AI, 2024" }, "probes": { "linear": { "0-10": { "index": 30.97, "ci": [ 30.15, 31.8 ], "rank": 20, "rank_ci": [ 19, 20 ], "index4": 30.49, "T1": 0.7045, "T2": 0.6779, "T2_labels": 10, "T4": 0.5823, "T1_n": 4878, "T4_n": 6077, "T3": 0.6453, "T3_n": 137, "T3_novel": null, "skills": { "T1": 40.89, "T2": 35.57, "T4": 16.45, "T3": 29.06 }, "T1_bins": { "zero": 0.8506, "1-10": 0.6852, "11-100": 0.6758, "101-1000": 0.6748, ">1000": 0.6459 }, "ci_family": [ 29.15, 33.06 ], "rank_ci_family": [ 19, 20 ], "index_r14": 30.97, "rank_r14": 20, "T4_recall10": 0.0038 }, "all": { "index": 29.3, "ci": [ 28.84, 29.77 ], "rank": 20, "rank_ci": [ 20, 20 ], "index4": 30.01, "T1": 0.6832, "T2": 0.6852, "T2_labels": 11, "T4": 0.5711, "T1_n": null, "T4_n": null, "T3": 0.6606, "T3_n": 347, "T3_novel": null, "skills": { "T1": 36.65, "T2": 37.04, "T4": 14.22, "T3": 32.13 }, "T4_recall10": 0.0031 } }, "mlp": { "0-10": { "index": 38.92, "ci": [ 38.29, 39.57 ], "rank": 20, "rank_ci": [ 20, 20 ], "index4": 44.46, "T1": 0.8409, "T2": 0.6606, "T2_labels": 10, "T4": 0.5823, "T1_n": 4878, "T4_n": 6077, "T3": 0.8054, "T3_n": 137, "T3_novel": null, "skills": { "T1": 68.19, "T2": 32.13, "T4": 16.45, "T3": 61.08 }, "T1_bins": { "zero": 0.9226, "1-10": 0.8302, "11-100": 0.8082, "101-1000": 0.7924, ">1000": 0.7675 }, "T4_recall10": 0.0038 }, "all": { "index": 36.33, "ci": [ 35.9, 36.74 ], "rank": 20, "rank_ci": [ 20, 20 ], "index4": 41.53, "T1": 0.8136, "T2": 0.6603, "T2_labels": 11, "T4": 0.5711, "T1_n": null, "T4_n": null, "T3": 0.7857, "T3_n": 347, "T3_novel": null, "skills": { "T1": 62.72, "T2": 32.05, "T4": 14.22, "T3": 57.14 }, "T4_recall10": 0.0031 } } }, "coverage": { "n_present": 19236, "share_missing": 0.001, "missing_by_bin": { "zero": 4, "1-10": 6, "11-100": 5, "101-1000": 4, ">1000": 1 } } }, { "track": "O", "tag": "rna1b6", "name": "GB.RNA-1.6B", "short": "GB.RNA-1.6B", "version": "v1", "in_v0": false, "modality": "rna", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 19228, "meta": { "training": "42 million non-coding RNA sequences, RNAcentral 24.0", "date": "2024", "signal": "none", "t3": "yes", "t3_why": "sequence rule", "source": "https://huggingface.co/genbio-ai/AIDO.RNA-1.6B", "cite": "GenBio AI, 2024" }, "probes": { "linear": { "0-10": { "index": 39.9, "ci": [ 39.2, 40.65 ], "rank": 12, "rank_ci": [ 12, 13 ], "index4": 42.58, "T1": 0.8003, "T2": 0.6707, "T2_labels": 10, "T4": 0.6275, "T1_n": 4878, "T4_n": 6077, "T3": 0.753, "T3_n": 137, "T3_novel": null, "skills": { "T1": 60.06, "T2": 34.15, "T4": 25.5, "T3": 50.6 }, "T1_bins": { "zero": 0.8955, "1-10": 0.7877, "11-100": 0.7842, "101-1000": 0.7706, ">1000": 0.7405 }, "ci_family": [ 35.47, 43.74 ], "rank_ci_family": [ 11, 14 ], "index_r14": 39.9, "rank_r14": 12, "T4_recall10": 0.0077 }, "all": { "index": 40.44, "ci": [ 39.99, 40.87 ], "rank": 15, "rank_ci": [ 14, 15 ], "index4": 44.42, "T1": 0.7852, "T2": 0.7085, "T2_labels": 11, "T4": 0.6129, "T1_n": null, "T4_n": null, "T3": 0.7818, "T3_n": 347, "T3_novel": null, "skills": { "T1": 57.03, "T2": 41.71, "T4": 22.57, "T3": 56.36 }, "T4_recall10": 0.0059 } }, "mlp": { "0-10": { "index": 47.27, "ci": [ 46.6, 47.93 ], "rank": 13, "rank_ci": [ 11, 13 ], "index4": 54.1, "T1": 0.887, "T2": 0.6945, "T2_labels": 10, "T4": 0.6275, "T1_n": 4878, "T4_n": 6077, "T3": 0.8729, "T3_n": 137, "T3_novel": null, "skills": { "T1": 77.4, "T2": 38.9, "T4": 25.5, "T3": 74.59 }, "T1_bins": { "zero": 0.9521, "1-10": 0.8784, "11-100": 0.8638, "101-1000": 0.852, ">1000": 0.824 }, "T4_recall10": 0.0077 }, "all": { "index": 45.95, "ci": [ 45.51, 46.34 ], "rank": 16, "rank_ci": [ 16, 17 ], "index4": 52.95, "T1": 0.8673, "T2": 0.7091, "T2_labels": 11, "T4": 0.6129, "T1_n": null, "T4_n": null, "T3": 0.8698, "T3_n": 347, "T3_novel": null, "skills": { "T1": 73.45, "T2": 41.83, "T4": 22.57, "T3": 73.96 }, "T4_recall10": 0.0059 } } }, "coverage": { "n_present": 19228, "share_missing": 0.0015, "missing_by_bin": { "zero": 6, "1-10": 8, "11-100": 9, "101-1000": 4, ">1000": 1 } } }, { "track": "O", "tag": "esm1b", "name": "ESM1b", "short": "ESM1b", "version": "v1", "in_v0": false, "modality": "protein", "cls": "bio", "text": null, "train_text": null, "encoder": null, "text_label": "No text embedded; trained without text", "gen_share": null, "n_genes": 19133, "meta": { "training": "UniRef50 2018_03, masked language modelling", "date": "2018", "signal": "none", "t3": "yes", "t3_why": "sequence rule", "source": "https://github.com/facebookresearch/esm", "cite": "Rives et al., 2021" }, "probes": { "linear": { "0-10": { "index": 43.91, "ci": [ 43.15, 44.71 ], "rank": 6, "rank_ci": [ 5, 6 ], "index4": 46.39, "T1": 0.8733, "T2": 0.6736, "T2_labels": 10, "T4": 0.6118, "T1_n": 4878, "T4_n": 6077, "T3": 0.7691, "T3_n": 137, "T3_novel": null, "skills": { "T1": 74.66, "T2": 34.72, "T4": 22.36, "T3": 53.83 }, "T1_bins": { "zero": 0.9556, "1-10": 0.8624, "11-100": 0.8528, "101-1000": 0.8268, ">1000": 0.7953 }, "ci_family": [ 39.08, 48.36 ], "rank_ci_family": [ 5, 6 ], "index_r14": 43.9, "rank_r14": 5, "T4_recall10": 0.0174 }, "all": { "index": 44.46, "ci": [ 44.03, 44.9 ], "rank": 10, "rank_ci": [ 10, 11 ], "index4": 49.6, "T1": 0.8523, "T2": 0.7138, "T2_labels": 11, "T4": 0.6007, "T1_n": null, "T4_n": null, "T3": 0.8252, "T3_n": 347, "T3_novel": null, "skills": { "T1": 70.45, "T2": 42.77, "T4": 20.15, "T3": 65.04 }, "T4_recall10": 0.0166 } }, "mlp": { "0-10": { "index": 53.25, "ci": [ 52.61, 53.91 ], "rank": 4, "rank_ci": [ 4, 4 ], "index4": 60.75, "T1": 0.9449, "T2": 0.742, "T2_labels": 10, "T4": 0.6118, "T1_n": 4878, "T4_n": 6077, "T3": 0.9164, "T3_n": 137, "T3_novel": null, "skills": { "T1": 88.99, "T2": 48.4, "T4": 22.36, "T3": 83.27 }, "T1_bins": { "zero": 0.9818, "1-10": 0.9401, "11-100": 0.928, "101-1000": 0.907, ">1000": 0.8773 }, "T4_recall10": 0.0174 }, "all": { "index": 52.24, "ci": [ 51.83, 52.64 ], "rank": 8, "rank_ci": [ 7, 9 ], "index4": 60.56, "T1": 0.9276, "T2": 0.7552, "T2_labels": 11, "T4": 0.6007, "T1_n": null, "T4_n": null, "T3": 0.9276, "T3_n": 347, "T3_novel": null, "skills": { "T1": 85.51, "T2": 51.05, "T4": 20.15, "T3": 85.51 }, "T4_recall10": 0.0166 } } }, "coverage": { "n_present": 19133, "share_missing": 0.0064, "missing_by_bin": { "zero": 16, "1-10": 59, "11-100": 42, "101-1000": 6, ">1000": 0 } } }, { "track": "O", "tag": "protst", "name": "ProtST-ESM1b (text-aligned)", "short": "ProtST-ESM1b", "version": "v1", "in_v0": false, "modality": "text-aligned protein", "cls": "biotext", "text": null, "train_text": "UniProt function text (Swiss-Prot 2022)", "encoder": null, "text_label": "No text embedded; UniProt function text (Swiss-Prot 2022) used in training", "gen_share": null, "n_genes": 19133, "meta": { "training": "ESM1b aligned with Swiss-Prot function text (ProtDescribe)", "date": "Swiss-Prot 2022", "signal": "UniProt function text (co-curated with GOA)", "t3": "yes", "t3_why": "training text predates the cutoff", "source": "https://github.com/DeepGraphLearning/ProtST", "cite": "Xu et al., 2023" }, "probes": { "linear": { "0-10": { "index": 46.33, "ci": [ 45.57, 47.05 ], "rank": 3, "rank_ci": [ 2, 3 ], "index4": 51.54, "T1": 0.8918, "T2": 0.7293, "T2_labels": 10, "T4": 0.5739, "T1_n": 4878, "T4_n": 6077, "T3": 0.8359, "T3_n": 137, "T3_novel": null, "skills": { "T1": 78.36, "T2": 45.86, "T4": 14.78, "T3": 67.18 }, "T1_bins": { "zero": 0.9543, "1-10": 0.8835, "11-100": 0.8805, "101-1000": 0.8567, ">1000": 0.8266 }, "ci_family": [ 42.41, 49.97 ], "rank_ci_family": [ 1, 4 ], "index_r14": 46.33, "rank_r14": 3, "T4_recall10": 0.0139 }, "all": { "index": 48.03, "ci": [ 47.59, 48.46 ], "rank": 3, "rank_ci": [ 3, 4 ], "index4": 54.29, "T1": 0.8778, "T2": 0.773, "T2_labels": 11, "T4": 0.5697, "T1_n": null, "T4_n": null, "T3": 0.8653, "T3_n": 347, "T3_novel": null, "skills": { "T1": 75.55, "T2": 54.6, "T4": 13.94, "T3": 73.05 }, "T4_recall10": 0.0128 } }, "mlp": { "0-10": { "index": 51.57, "ci": [ 50.94, 52.22 ], "rank": 8, "rank_ci": [ 6, 8 ], "index4": 59.67, "T1": 0.9493, "T2": 0.7504, "T2_labels": 10, "T4": 0.5739, "T1_n": 4878, "T4_n": 6077, "T3": 0.9199, "T3_n": 137, "T3_novel": null, "skills": { "T1": 89.86, "T2": 50.08, "T4": 14.78, "T3": 83.98 }, "T1_bins": { "zero": 0.9804, "1-10": 0.9452, "11-100": 0.938, "101-1000": 0.9188, ">1000": 0.8965 }, "T4_recall10": 0.0139 }, "all": { "index": 51.76, "ci": [ 51.33, 52.13 ], "rank": 10, "rank_ci": [ 9, 10 ], "index4": 60.27, "T1": 0.9365, "T2": 0.7702, "T2_labels": 11, "T4": 0.5697, "T1_n": null, "T4_n": null, "T3": 0.929, "T3_n": 347, "T3_novel": null, "skills": { "T1": 87.3, "T2": 54.04, "T4": 13.94, "T3": 85.8 }, "T4_recall10": 0.0128 } } }, "coverage": { "n_present": 19133, "share_missing": 0.0064, "missing_by_bin": { "zero": 16, "1-10": 59, "11-100": 42, "101-1000": 6, ">1000": 0 } } }, { "track": "O", "tag": "protrek", "name": "ProTrek-650M (text-aligned)", "short": "ProTrek-650M", "version": "v1", "in_v0": false, "modality": "text-aligned protein", "cls": "biotext", "text": null, "train_text": "UniProt function text (Swiss-Prot 2024)", "encoder": null, "text_label": "No text embedded; UniProt function text (Swiss-Prot 2024) used in training", "gen_share": null, "n_genes": 19133, "meta": { "training": "ESM2-650M aligned with Swiss-Prot text and structure, contrastive", "date": "Swiss-Prot 2024", "signal": "UniProt function text", "t3": "no", "t3_why": "training text after the cutoff", "source": "https://github.com/westlake-repl/ProTrek", "cite": "Su et al., 2024" }, "probes": { "linear": { "0-10": { "index": 46.96, "ci": [ 46.21, 47.65 ], "rank": 2, "rank_ci": [ 1, 2 ], "index4": null, "T1": 0.9037, "T2": 0.6979, "T2_labels": 10, "T4": 0.6029, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 80.74, "T2": 39.57, "T4": 20.58, "T3": null }, "T1_bins": { "zero": 0.952, "1-10": 0.8973, "11-100": 0.8908, "101-1000": 0.876, ">1000": 0.8597 }, "ci_family": [ 41.29, 51.72 ], "rank_ci_family": [ 1, 3 ], "index_r14": 46.96, "rank_r14": 2, "T4_recall10": 0.0173 }, "all": { "index": 48.77, "ci": [ 48.33, 49.2 ], "rank": 2, "rank_ci": [ 2, 2 ], "index4": null, "T1": 0.8909, "T2": 0.7486, "T2_labels": 11, "T4": 0.5921, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 78.18, "T2": 49.72, "T4": 18.41, "T3": null }, "T4_recall10": 0.016 } }, "mlp": { "0-10": { "index": 53.94, "ci": [ 53.29, 54.56 ], "rank": 3, "rank_ci": [ 2, 3 ], "index4": null, "T1": 0.9596, "T2": 0.7467, "T2_labels": 10, "T4": 0.6029, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 91.91, "T2": 49.33, "T4": 20.58, "T3": null }, "T1_bins": { "zero": 0.9858, "1-10": 0.9561, "11-100": 0.9438, "101-1000": 0.9219, ">1000": 0.8972 }, "T4_recall10": 0.0173 }, "all": { "index": 54.03, "ci": [ 53.6, 54.44 ], "rank": 3, "rank_ci": [ 3, 4 ], "index4": null, "T1": 0.943, "T2": 0.7754, "T2_labels": 11, "T4": 0.5921, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 88.59, "T2": 55.09, "T4": 18.41, "T3": null }, "T4_recall10": 0.016 } } }, "coverage": { "n_present": 19133, "share_missing": 0.0064, "missing_by_bin": { "zero": 16, "1-10": 59, "11-100": 42, "101-1000": 6, ">1000": 0 } } }, { "track": "O", "tag": "text_uniprot", "name": "UniProt function text only", "short": "UniProt function text", "version": "v1", "in_v0": false, "modality": "text", "cls": "clean", "text": "uniprot", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "UniProt function text only (co-curated with GOA)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "UniProt function text of October 2026, embedded with Qwen3-Embedding-8B", "date": "text 2026; encoder 2025", "signal": "UniProt text (co-curated with GOA)", "t3": "no", "t3_why": "text after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 42.46, "ci": [ 41.66, 43.19 ], "rank": 9, "rank_ci": [ 8, 9 ], "index4": null, "T1": 0.9162, "T2": 0.6466, "T2_labels": 10, "T4": 0.574, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.24, "T2": 29.32, "T4": 14.81, "T3": null }, "T1_bins": { "zero": 0.9398, "1-10": 0.9131, "11-100": 0.9109, "101-1000": 0.889, ">1000": 0.8643 }, "T2_per_label": { "depmap_essential": 0.8151, "hpa_loc_nucleoplasm": 0.6546, "hpa_loc_cytosol": 0.54, "hpa_loc_plasma_membrane": 0.5634, "hpa_loc_vesicles": 0.5203, "hpa_loc_mitochondria": 0.6872, "hpa_loc_golgi_apparatus": 0.5535, "gtex_tissue_specific": 0.7443, "hpa_tissue_enriched": 0.7576, "constraint": 0.6303 }, "T2_per_label_source": "round-14 regeneration", "ci_family": [ 38.0, 46.25 ], "rank_ci_family": [ 7, 10 ], "index_r14": 42.46, "rank_r14": 9, "T4_recall10": 0.0204 }, "all": { "index": 46.56, "ci": [ 46.13, 47.05 ], "rank": 6, "rank_ci": [ 5, 7 ], "index4": null, "T1": 0.907, "T2": 0.7064, "T2_labels": 11, "T4": 0.5851, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 81.4, "T2": 41.28, "T4": 17.02, "T3": null }, "T2_per_label": { "depmap_essential": 0.8557, "hpa_loc_nucleoplasm": 0.704, "hpa_loc_cytosol": 0.5986, "hpa_loc_plasma_membrane": 0.6713, "hpa_loc_vesicles": 0.6037, "hpa_loc_mitochondria": 0.7567, "hpa_loc_golgi_apparatus": 0.6074, "gtex_tissue_specific": 0.8253, "hpa_tissue_enriched": 0.8108, "clingen_hi": 0.6294, "constraint": 0.7073 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0312 } }, "mlp": { "0-10": { "index": 51.61, "ci": [ 50.89, 52.29 ], "rank": 7, "rank_ci": [ 6, 8 ], "index4": null, "T1": 0.9668, "T2": 0.7333, "T2_labels": 10, "T4": 0.574, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 93.36, "T2": 46.65, "T4": 14.81, "T3": null }, "T1_bins": { "zero": 0.9844, "1-10": 0.9645, "11-100": 0.9506, "101-1000": 0.9244, ">1000": 0.8938 }, "T2_per_label": { "depmap_essential": 0.9399, "hpa_loc_nucleoplasm": 0.6621, "hpa_loc_cytosol": 0.6215, "hpa_loc_plasma_membrane": 0.6643, "hpa_loc_vesicles": 0.6467, "hpa_loc_mitochondria": 0.75, "hpa_loc_golgi_apparatus": 0.6972, "gtex_tissue_specific": 0.8293, "hpa_tissue_enriched": 0.8289, "constraint": 0.6927 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0204 }, "all": { "index": 53.98, "ci": [ 53.55, 54.39 ], "rank": 4, "rank_ci": [ 3, 5 ], "index4": null, "T1": 0.9488, "T2": 0.7758, "T2_labels": 11, "T4": 0.5851, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 89.75, "T2": 55.16, "T4": 17.02, "T3": null }, "T2_per_label": { "depmap_essential": 0.9565, "hpa_loc_nucleoplasm": 0.7386, "hpa_loc_cytosol": 0.6559, "hpa_loc_plasma_membrane": 0.7708, "hpa_loc_vesicles": 0.7004, "hpa_loc_mitochondria": 0.832, "hpa_loc_golgi_apparatus": 0.7294, "gtex_tissue_specific": 0.8761, "hpa_tissue_enriched": 0.8605, "clingen_hi": 0.6434, "constraint": 0.7701 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0312 } } }, "coverage": { "n_present": 19256, "share_missing": 0.0, "missing_by_bin": { "zero": 0, "1-10": 0, "11-100": 0, "101-1000": 0, ">1000": 0 } } }, { "track": "O", "tag": "fusion_p16b_refseq", "name": "GB.Protein-16B ⊕ rich summary-filtered text", "short": "GB.Protein-16B ⊕ text", "version": "v1", "in_v0": false, "modality": "mixture", "cls": "biotext", "text": "mixture", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Declared mixture; the text component is summary-filtered text (contains UniProt function text)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "concatenation of the standardised GB.Protein-16B and rich summary-filtered text embeddings", "date": "mixed", "signal": "UniProt text", "t3": "no", "t3_why": "text component after the cutoff", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 47.09, "ci": [ 46.32, 47.81 ], "rank": 1, "rank_ci": [ 1, 2 ], "index4": null, "T1": 0.9212, "T2": 0.6578, "T2_labels": 10, "T4": 0.6273, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.24, "T2": 31.57, "T4": 25.45, "T3": null }, "T1_bins": { "zero": 0.9491, "1-10": 0.9175, "11-100": 0.9117, "101-1000": 0.8859, ">1000": 0.8603 }, "T2_per_label": { "depmap_essential": 0.8147, "hpa_loc_nucleoplasm": 0.6455, "hpa_loc_cytosol": 0.5316, "hpa_loc_plasma_membrane": 0.5563, "hpa_loc_vesicles": 0.5266, "hpa_loc_mitochondria": 0.7074, "hpa_loc_golgi_apparatus": 0.528, "gtex_tissue_specific": 0.7938, "hpa_tissue_enriched": 0.7817, "constraint": 0.6929 }, "T2_per_label_source": "round-14 regeneration", "ci_family": [ 41.75, 51.41 ], "rank_ci_family": [ 1, 3 ], "index_r14": 47.09, "rank_r14": 1, "T4_recall10": 0.0238 }, "all": { "index": 50.43, "ci": [ 50.0, 50.87 ], "rank": 1, "rank_ci": [ 1, 1 ], "index4": null, "T1": 0.9081, "T2": 0.7203, "T2_labels": 11, "T4": 0.628, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 81.63, "T2": 44.05, "T4": 25.6, "T3": null }, "T2_per_label": { "depmap_essential": 0.8457, "hpa_loc_nucleoplasm": 0.7083, "hpa_loc_cytosol": 0.5943, "hpa_loc_plasma_membrane": 0.6618, "hpa_loc_vesicles": 0.6026, "hpa_loc_mitochondria": 0.7749, "hpa_loc_golgi_apparatus": 0.59, "gtex_tissue_specific": 0.8633, "hpa_tissue_enriched": 0.8351, "clingen_hi": 0.6895, "constraint": 0.7575 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0322 } }, "mlp": { "0-10": { "index": 58.25, "ci": [ 57.57, 58.86 ], "rank": 1, "rank_ci": [ 1, 1 ], "index4": null, "T1": 0.9696, "T2": 0.7769, "T2_labels": 10, "T4": 0.6273, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 93.91, "T2": 55.38, "T4": 25.45, "T3": null }, "T1_bins": { "zero": 0.986, "1-10": 0.9674, "11-100": 0.9542, "101-1000": 0.9293, ">1000": 0.8993 }, "T2_per_label": { "depmap_essential": 0.9602, "hpa_loc_nucleoplasm": 0.6625, "hpa_loc_cytosol": 0.6421, "hpa_loc_plasma_membrane": 0.7012, "hpa_loc_vesicles": 0.6593, "hpa_loc_mitochondria": 0.7844, "hpa_loc_golgi_apparatus": 0.7047, "gtex_tissue_specific": 0.9019, "hpa_tissue_enriched": 0.8897, "constraint": 0.8629 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0238 }, "all": { "index": 59.51, "ci": [ 59.13, 59.89 ], "rank": 1, "rank_ci": [ 1, 1 ], "index4": null, "T1": 0.9524, "T2": 0.8122, "T2_labels": 11, "T4": 0.628, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 90.48, "T2": 62.44, "T4": 25.6, "T3": null }, "T2_per_label": { "depmap_essential": 0.9659, "hpa_loc_nucleoplasm": 0.7625, "hpa_loc_cytosol": 0.6882, "hpa_loc_plasma_membrane": 0.7873, "hpa_loc_vesicles": 0.7153, "hpa_loc_mitochondria": 0.8524, "hpa_loc_golgi_apparatus": 0.7409, "gtex_tissue_specific": 0.9211, "hpa_tissue_enriched": 0.9003, "clingen_hi": 0.7221, "constraint": 0.8783 }, "T2_per_label_source": "round-14 regeneration", "T4_recall10": 0.0322 } } }, "coverage": { "n_present": 19133, "share_missing": 0.0064, "missing_by_bin": { "zero": 16, "1-10": 59, "11-100": 42, "101-1000": 6, ">1000": 0 } } }, { "track": "O", "tag": "text_refseq_nouni", "name": "Rich summary-filtered text without UniProt function text", "short": "Summary-filtered, no UniProt text", "version": "v1.1", "in_v0": false, "modality": "text", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Summary-filtered text (generated summaries removed); UniProt function text removed as well", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "“Gene SYMBOL (name). RefSeq summary” without UniProt function text, Qwen3-Embedding-8B", "date": "text 2026; encoder 2025", "signal": "RefSeq prose; encoder memory", "t3": "no", "t3_why": "text after the cutoff", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "this work (round 14)" }, "probes": { "linear": { "0-10": { "index": 37.85, "ci": [ 37.01, 38.64 ], "rank": null, "rank_ci": null, "index4": null, "T1": 0.8595, "T2": 0.64, "T2_labels": 10, "T4": 0.5683, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 71.9, "T2": 28.0, "T4": 13.65, "T3": null }, "T2_per_label": { "depmap_essential": 0.8143, "hpa_loc_nucleoplasm": 0.6135, "hpa_loc_cytosol": 0.5396, "hpa_loc_plasma_membrane": 0.5544, "hpa_loc_vesicles": 0.5284, "hpa_loc_mitochondria": 0.695, "hpa_loc_golgi_apparatus": 0.5569, "gtex_tissue_specific": 0.722, "hpa_tissue_enriched": 0.7462, "constraint": 0.6299 }, "T1_bins": { "zero": 0.9361, "1-10": 0.8494, "11-100": 0.8653, "101-1000": 0.8587, ">1000": 0.8433 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0168 }, "all": { "index": 42.56, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8618, "T2": 0.7094, "T2_labels": 11, "T4": 0.5673, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 72.35, "T2": 41.88, "T4": 13.46, "T3": null }, "T2_fixed": 0.7089, "index_fixedT2": 42.53, "T2_per_label": { "depmap_essential": 0.8335, "hpa_loc_nucleoplasm": 0.694, "hpa_loc_cytosol": 0.601, "hpa_loc_plasma_membrane": 0.665, "hpa_loc_vesicles": 0.6091, "hpa_loc_mitochondria": 0.7693, "hpa_loc_golgi_apparatus": 0.6085, "gtex_tissue_specific": 0.8069, "hpa_tissue_enriched": 0.7934, "clingen_hi": 0.7138, "constraint": 0.7087 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0239 } }, "mlp": { "0-10": { "index": 46.92, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9334, "T2": 0.7022, "T2_labels": 10, "T4": 0.5683, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 86.68, "T2": 40.43, "T4": 13.65, "T3": null }, "T2_per_label": { "depmap_essential": 0.8986, "hpa_loc_nucleoplasm": 0.6278, "hpa_loc_cytosol": 0.5846, "hpa_loc_plasma_membrane": 0.6645, "hpa_loc_vesicles": 0.6293, "hpa_loc_mitochondria": 0.7486, "hpa_loc_golgi_apparatus": 0.6433, "gtex_tissue_specific": 0.7637, "hpa_tissue_enriched": 0.7919, "constraint": 0.6694 }, "T1_bins": { "zero": 0.9751, "1-10": 0.9279, "11-100": 0.9238, "101-1000": 0.9064, ">1000": 0.8768 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0168 }, "all": { "index": 50.31, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.922, "T2": 0.7653, "T2_labels": 11, "T4": 0.5673, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.4, "T2": 53.06, "T4": 13.46, "T3": null }, "T2_fixed": 0.7668, "index_fixedT2": 50.41, "T2_per_label": { "depmap_essential": 0.9245, "hpa_loc_nucleoplasm": 0.7223, "hpa_loc_cytosol": 0.6428, "hpa_loc_plasma_membrane": 0.7585, "hpa_loc_vesicles": 0.6753, "hpa_loc_mitochondria": 0.8193, "hpa_loc_golgi_apparatus": 0.6917, "gtex_tissue_specific": 0.8458, "hpa_tissue_enriched": 0.8356, "clingen_hi": 0.7501, "constraint": 0.7524 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0239 } } }, "t3_only": false }, { "track": "O", "tag": "fusion_p16b_refseq_nouni", "name": "GB.Protein-16B ⊕ summary-filtered text without UniProt text", "short": "GB.Protein-16B ⊕ text, no UniProt", "version": "v1.1", "in_v0": false, "modality": "mixture", "cls": "biotext", "text": "mixture", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Declared mixture; the text component is summary-filtered text without UniProt function text", "gen_share": 0.0, "n_genes": 19133, "meta": { "training": "concatenation of GB.Protein-16B and the text row without UniProt function text", "date": "mixed", "signal": "RefSeq prose", "t3": "no", "t3_why": "text component after the cutoff", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work (round 14)" }, "probes": { "linear": { "0-10": { "index": 42.88, "ci": [ 42.12, 43.65 ], "rank": null, "rank_ci": null, "index4": null, "T1": 0.8741, "T2": 0.6502, "T2_labels": 10, "T4": 0.619, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 74.81, "T2": 30.04, "T4": 23.8, "T3": null }, "T2_per_label": { "depmap_essential": 0.8033, "hpa_loc_nucleoplasm": 0.6333, "hpa_loc_cytosol": 0.5401, "hpa_loc_plasma_membrane": 0.5268, "hpa_loc_vesicles": 0.5146, "hpa_loc_mitochondria": 0.6973, "hpa_loc_golgi_apparatus": 0.5361, "gtex_tissue_specific": 0.7777, "hpa_tissue_enriched": 0.7657, "constraint": 0.7073 }, "T1_bins": { "zero": 0.9416, "1-10": 0.8652, "11-100": 0.8697, "101-1000": 0.8536, ">1000": 0.8347 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0206 }, "all": { "index": 46.08, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8669, "T2": 0.7112, "T2_labels": 11, "T4": 0.6131, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 73.39, "T2": 42.24, "T4": 22.61, "T3": null }, "T2_fixed": 0.7112, "index_fixedT2": 46.08, "T2_per_label": { "depmap_essential": 0.8212, "hpa_loc_nucleoplasm": 0.6971, "hpa_loc_cytosol": 0.5962, "hpa_loc_plasma_membrane": 0.6421, "hpa_loc_vesicles": 0.5914, "hpa_loc_mitochondria": 0.7618, "hpa_loc_golgi_apparatus": 0.5779, "gtex_tissue_specific": 0.8476, "hpa_tissue_enriched": 0.8176, "clingen_hi": 0.7114, "constraint": 0.7587 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0267 } }, "mlp": { "0-10": { "index": 54.89, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9462, "T2": 0.7581, "T2_labels": 10, "T4": 0.619, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 89.24, "T2": 51.63, "T4": 23.8, "T3": null }, "T2_per_label": { "depmap_essential": 0.9592, "hpa_loc_nucleoplasm": 0.6286, "hpa_loc_cytosol": 0.6321, "hpa_loc_plasma_membrane": 0.6804, "hpa_loc_vesicles": 0.648, "hpa_loc_mitochondria": 0.7803, "hpa_loc_golgi_apparatus": 0.6765, "gtex_tissue_specific": 0.8752, "hpa_tissue_enriched": 0.8452, "constraint": 0.856 }, "T1_bins": { "zero": 0.9805, "1-10": 0.9417, "11-100": 0.9329, "101-1000": 0.9112, ">1000": 0.883 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0206 }, "all": { "index": 56.36, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9313, "T2": 0.8011, "T2_labels": 11, "T4": 0.6131, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 86.25, "T2": 60.21, "T4": 22.61, "T3": null }, "T2_fixed": 0.8066, "index_fixedT2": 56.73, "T2_per_label": { "depmap_essential": 0.9604, "hpa_loc_nucleoplasm": 0.7406, "hpa_loc_cytosol": 0.6742, "hpa_loc_plasma_membrane": 0.7718, "hpa_loc_vesicles": 0.7018, "hpa_loc_mitochondria": 0.8446, "hpa_loc_golgi_apparatus": 0.7198, "gtex_tissue_specific": 0.9083, "hpa_tissue_enriched": 0.874, "clingen_hi": 0.7462, "constraint": 0.8701 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0267 } } }, "t3_only": false }, { "track": "O", "tag": "fusion_esm2_refseq", "name": "ESM2-650M ⊕ rich summary-filtered text", "short": "ESM2-650M ⊕ text", "version": "v1.1", "in_v0": false, "modality": "mixture", "cls": "biotext", "text": "mixture", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Declared mixture; the text component is summary-filtered text (contains UniProt function text)", "gen_share": 0.0, "n_genes": 19133, "meta": { "training": "concatenation of re-embedded ESM2-650M (round 14) and rich summary-filtered text", "date": "mixed", "signal": "UniProt text", "t3": "no", "t3_why": "text component after the cutoff", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work (round 14)" }, "probes": { "linear": { "0-10": { "index": 45.49, "ci": [ 44.8, 46.21 ], "rank": null, "rank_ci": null, "index4": null, "T1": 0.9211, "T2": 0.6611, "T2_labels": 10, "T4": 0.6002, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.21, "T2": 32.22, "T4": 20.04, "T3": null }, "T2_per_label": { "depmap_essential": 0.8207, "hpa_loc_nucleoplasm": 0.6472, "hpa_loc_cytosol": 0.5432, "hpa_loc_plasma_membrane": 0.5891, "hpa_loc_vesicles": 0.5231, "hpa_loc_mitochondria": 0.7002, "hpa_loc_golgi_apparatus": 0.5408, "gtex_tissue_specific": 0.7811, "hpa_tissue_enriched": 0.788, "constraint": 0.6777 }, "T1_bins": { "zero": 0.955, "1-10": 0.9166, "11-100": 0.9127, "101-1000": 0.8878, ">1000": 0.865 }, "T2_per_label_source": "round 14", "T4_recall10": 0.022 }, "all": { "index": 49.28, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9091, "T2": 0.723, "T2_labels": 11, "T4": 0.6071, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 81.81, "T2": 44.61, "T4": 21.41, "T3": null }, "T2_fixed": 0.7254, "index_fixedT2": 49.43, "T2_per_label": { "depmap_essential": 0.8497, "hpa_loc_nucleoplasm": 0.7076, "hpa_loc_cytosol": 0.6065, "hpa_loc_plasma_membrane": 0.6763, "hpa_loc_vesicles": 0.6074, "hpa_loc_mitochondria": 0.7781, "hpa_loc_golgi_apparatus": 0.6016, "gtex_tissue_specific": 0.854, "hpa_tissue_enriched": 0.8316, "clingen_hi": 0.6996, "constraint": 0.7411 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0305 } }, "mlp": { "0-10": { "index": 55.38, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9684, "T2": 0.7621, "T2_labels": 10, "T4": 0.6002, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 93.69, "T2": 52.41, "T4": 20.04, "T3": null }, "T2_per_label": { "depmap_essential": 0.9593, "hpa_loc_nucleoplasm": 0.659, "hpa_loc_cytosol": 0.6278, "hpa_loc_plasma_membrane": 0.696, "hpa_loc_vesicles": 0.6599, "hpa_loc_mitochondria": 0.7943, "hpa_loc_golgi_apparatus": 0.7069, "gtex_tissue_specific": 0.8523, "hpa_tissue_enriched": 0.8623, "constraint": 0.8028 }, "T1_bins": { "zero": 0.9837, "1-10": 0.9664, "11-100": 0.9524, "101-1000": 0.9269, ">1000": 0.8921 }, "T2_per_label_source": "round 14", "T4_recall10": 0.022 }, "all": { "index": 57.64, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9506, "T2": 0.807, "T2_labels": 11, "T4": 0.6071, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 90.12, "T2": 61.39, "T4": 21.41, "T3": null }, "T2_fixed": 0.8137, "index_fixedT2": 58.09, "T2_per_label": { "depmap_essential": 0.9671, "hpa_loc_nucleoplasm": 0.7581, "hpa_loc_cytosol": 0.6806, "hpa_loc_plasma_membrane": 0.7854, "hpa_loc_vesicles": 0.7113, "hpa_loc_mitochondria": 0.8554, "hpa_loc_golgi_apparatus": 0.7513, "gtex_tissue_specific": 0.9013, "hpa_tissue_enriched": 0.8835, "clingen_hi": 0.7396, "constraint": 0.843 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0305 } } }, "t3_only": false }, { "track": "O", "tag": "fusion_esm2_refseq_nouni", "name": "ESM2-650M ⊕ summary-filtered text without UniProt text", "short": "ESM2-650M ⊕ text, no UniProt", "version": "v1.1", "in_v0": false, "modality": "mixture", "cls": "biotext", "text": "mixture", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Declared mixture; the text component is summary-filtered text without UniProt function text", "gen_share": 0.0, "n_genes": 19133, "meta": { "training": "concatenation of re-embedded ESM2-650M (round 14) and the text row without UniProt function text", "date": "mixed", "signal": "RefSeq prose", "t3": "no", "t3_why": "text component after the cutoff", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work (round 14)" }, "probes": { "linear": { "0-10": { "index": 41.22, "ci": [ 40.45, 41.95 ], "rank": null, "rank_ci": null, "index4": null, "T1": 0.8734, "T2": 0.6526, "T2_labels": 10, "T4": 0.5923, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 74.68, "T2": 30.51, "T4": 18.46, "T3": null }, "T2_per_label": { "depmap_essential": 0.8126, "hpa_loc_nucleoplasm": 0.6334, "hpa_loc_cytosol": 0.5508, "hpa_loc_plasma_membrane": 0.5552, "hpa_loc_vesicles": 0.5316, "hpa_loc_mitochondria": 0.7046, "hpa_loc_golgi_apparatus": 0.5265, "gtex_tissue_specific": 0.759, "hpa_tissue_enriched": 0.7626, "constraint": 0.6894 }, "T1_bins": { "zero": 0.9481, "1-10": 0.8636, "11-100": 0.8701, "101-1000": 0.8558, ">1000": 0.8428 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0188 }, "all": { "index": 44.73, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8676, "T2": 0.7135, "T2_labels": 11, "T4": 0.5899, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 73.52, "T2": 42.7, "T4": 17.97, "T3": null }, "T2_fixed": 0.7127, "index_fixedT2": 44.68, "T2_per_label": { "depmap_essential": 0.8194, "hpa_loc_nucleoplasm": 0.7034, "hpa_loc_cytosol": 0.6009, "hpa_loc_plasma_membrane": 0.6534, "hpa_loc_vesicles": 0.6074, "hpa_loc_mitochondria": 0.7721, "hpa_loc_golgi_apparatus": 0.5862, "gtex_tissue_specific": 0.8364, "hpa_tissue_enriched": 0.8094, "clingen_hi": 0.7216, "constraint": 0.7384 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0249 } }, "mlp": { "0-10": { "index": 51.72, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9425, "T2": 0.741, "T2_labels": 10, "T4": 0.5923, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 88.5, "T2": 48.2, "T4": 18.46, "T3": null }, "T2_per_label": { "depmap_essential": 0.9446, "hpa_loc_nucleoplasm": 0.6324, "hpa_loc_cytosol": 0.619, "hpa_loc_plasma_membrane": 0.6807, "hpa_loc_vesicles": 0.6446, "hpa_loc_mitochondria": 0.7916, "hpa_loc_golgi_apparatus": 0.6686, "gtex_tissue_specific": 0.805, "hpa_tissue_enriched": 0.8129, "constraint": 0.8109 }, "T1_bins": { "zero": 0.9799, "1-10": 0.9376, "11-100": 0.9299, "101-1000": 0.9078, ">1000": 0.8787 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0188 }, "all": { "index": 54.03, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.928, "T2": 0.7926, "T2_labels": 11, "T4": 0.5899, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 85.59, "T2": 58.53, "T4": 17.97, "T3": null }, "T2_fixed": 0.7969, "index_fixedT2": 54.31, "T2_per_label": { "depmap_essential": 0.9516, "hpa_loc_nucleoplasm": 0.7388, "hpa_loc_cytosol": 0.666, "hpa_loc_plasma_membrane": 0.7734, "hpa_loc_vesicles": 0.698, "hpa_loc_mitochondria": 0.8439, "hpa_loc_golgi_apparatus": 0.7251, "gtex_tissue_specific": 0.8767, "hpa_tissue_enriched": 0.8555, "clingen_hi": 0.7504, "constraint": 0.8396 }, "T2_per_label_source": "round 14", "T4_recall10": 0.0249 } } }, "t3_only": false }, { "track": "O", "tag": "go23vec", "name": "GO annotation vector at the 2023 cutoff (non-IEA)", "short": "GO annotation vector (2023)", "version": "v1.1", "in_v0": false, "modality": "annotation vector", "cls": "annot", "text": null, "train_text": null, "encoder": null, "text_label": "Sequence embedding concatenated with the gene's own 2023 GO annotations (reference baseline)", "gen_share": null, "n_genes": 19256, "meta": { "training": "binary vector of the gene's GO annotations at the 2023-01-01 cutoff (non-IEA; 4,894 terms); a T3 reference, not a representation of the gene", "date": "2023", "signal": "the annotations themselves", "t3": "yes", "t3_why": "T3-only reference baseline (annotation-access rule)", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work (round 14)" }, "probes": { "linear": { "0-10": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": 0, "T4_n": 0, "T3": 0.4928, "T3_n": 137, "T3_novel": 0.3504, "skills": { "T1": null, "T2": null, "T4": null, "T3": -1.45 }, "T1_bins": { "zero": null, "1-10": null, "11-100": null, "101-1000": null, ">1000": null } }, "all": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": null, "T4_n": null, "T3": 0.6027, "T3_n": 347, "T3_novel": 0.3991, "skills": { "T1": null, "T2": null, "T4": null, "T3": 20.54 } } }, "mlp": { "0-10": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": 0, "T4_n": 0, "T3": 0.8892, "T3_n": 137, "T3_novel": 0.86, "skills": { "T1": null, "T2": null, "T4": null, "T3": 77.83 }, "T1_bins": { "zero": null, "1-10": null, "11-100": null, "101-1000": null, ">1000": null } }, "all": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": null, "T4_n": null, "T3": 0.9105, "T3_n": 347, "T3_novel": 0.8601, "skills": { "T1": null, "T2": null, "T4": null, "T3": 82.1 } } } }, "t3_only": true }, { "track": "O", "tag": "fusion_p16b_go23", "name": "GB.Protein-16B ⊕ GO annotation vector at the cutoff", "short": "GB.Protein-16B ⊕ GO vector", "version": "v1.1", "in_v0": false, "modality": "mixture", "cls": "biotext", "text": "mixture", "train_text": null, "encoder": null, "text_label": "Sequence embedding concatenated with the gene's own 2023 GO annotations (reference baseline)", "gen_share": 0.0, "n_genes": 19133, "meta": { "training": "concatenation of GB.Protein-16B and the 2023 GO annotation vector", "date": "mixed", "signal": "the annotations themselves", "t3": "yes", "t3_why": "T3-only reference", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work (round 14)" }, "probes": { "linear": { "0-10": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": 0, "T4_n": 0, "T3": 0.54, "T3_n": 137, "T3_novel": 0.4142, "skills": { "T1": null, "T2": null, "T4": null, "T3": 7.99 }, "T1_bins": { "zero": null, "1-10": null, "11-100": null, "101-1000": null, ">1000": null } }, "all": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": null, "T4_n": null, "T3": 0.633, "T3_n": 347, "T3_novel": 0.4453, "skills": { "T1": null, "T2": null, "T4": null, "T3": 26.59 } } }, "mlp": { "0-10": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": 0, "T4_n": 0, "T3": 0.8978, "T3_n": 137, "T3_novel": 0.8675, "skills": { "T1": null, "T2": null, "T4": null, "T3": 79.56 }, "T1_bins": { "zero": null, "1-10": null, "11-100": null, "101-1000": null, ">1000": null } }, "all": { "index": null, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": null, "T2": null, "T2_labels": 0, "T4": null, "T1_n": null, "T4_n": null, "T3": 0.918, "T3_n": 347, "T3_novel": 0.8685, "skills": { "T1": null, "T2": null, "T4": null, "T3": 83.6 } } } }, "t3_only": true }, { "track": "F", "tag": "t23l__qwen8b", "name": "Qwen3-Embedding-8B", "short": "Qwen3-Embedding-8B", "version": "v1", "in_v0": true, "modality": "text embedder", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "Qwen3-Embedding-8B", "text_label": "Scored on corpus L (GenePT's 2023 text, generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "Qwen3 language model tuned for embedding on web-scale text pairs", "date": "2025", "signal": "LLM memory", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-8B", "cite": "Zhang et al., 2025" }, "probes": { "linear": { "0-10": { "index": 34.19, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8325, "T2": 0.6242, "T2_labels": 10, "T4": 0.5561, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 66.49, "T2": 24.85, "T4": 11.22, "T3": null }, "T1_bins": { "zero": 0.9349, "1-10": 0.819, "11-100": 0.8551, "101-1000": 0.8591, ">1000": 0.8427 }, "S": 0.0779, "S_ci": [ 0.0719, 0.0841 ], "S_n": 4310, "T1_asfound_1-10": 0.8969 }, "all": { "index": 40.7, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.849, "T2": 0.7085, "T2_labels": 11, "T4": 0.5531, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 69.79, "T2": 41.69, "T4": 10.61, "T3": null }, "S": 0.0779, "S_ci": [ 0.0719, 0.0841 ], "S_n": 4310, "T1_asfound_1-10": 0.8969 } }, "mlp": { "0-10": { "index": 43.95, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9245, "T2": 0.6786, "T2_labels": 10, "T4": 0.5561, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.91, "T2": 35.72, "T4": 11.22, "T3": null }, "T1_bins": { "zero": 0.9721, "1-10": 0.9183, "11-100": 0.9231, "101-1000": 0.9117, ">1000": 0.8812 }, "S": 0.0334, "S_ci": [ 0.0301, 0.0366 ], "S_n": 4310, "T1_asfound_1-10": 0.9517 }, "all": { "index": 48.64, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9202, "T2": 0.7562, "T2_labels": 11, "T4": 0.5531, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.05, "T2": 51.25, "T4": 10.61, "T3": null }, "S": 0.0334, "S_ci": [ 0.0301, 0.0366 ], "S_n": 4310, "T1_asfound_1-10": 0.9517 } } } }, { "track": "F", "tag": "t23l__qwen06b", "name": "Qwen3-Embedding-0.6B", "short": "Qwen3-Embedding-0.6B", "version": "v1", "in_v0": true, "modality": "text embedder", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "Qwen3-Embedding-0.6B", "text_label": "Scored on corpus L (GenePT's 2023 text, generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "Qwen3 language model tuned for embedding on web-scale text pairs", "date": "2025", "signal": "LLM memory", "source": "https://huggingface.co/Qwen/Qwen3-Embedding-0.6B", "cite": "Zhang et al., 2025" }, "probes": { "linear": { "0-10": { "index": 32.93, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8218, "T2": 0.6206, "T2_labels": 10, "T4": 0.5515, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 64.36, "T2": 24.11, "T4": 10.3, "T3": null }, "T1_bins": { "zero": 0.9122, "1-10": 0.8099, "11-100": 0.8441, "101-1000": 0.8479, ">1000": 0.8382 }, "S": 0.0875, "S_ci": [ 0.0817, 0.0931 ], "S_n": 4310, "T1_asfound_1-10": 0.8974 }, "all": { "index": 38.71, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8382, "T2": 0.698, "T2_labels": 11, "T4": 0.5445, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 67.63, "T2": 39.6, "T4": 8.91, "T3": null }, "S": 0.0875, "S_ci": [ 0.0817, 0.0931 ], "S_n": 4310, "T1_asfound_1-10": 0.8974 } }, "mlp": { "0-10": { "index": 41.01, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9199, "T2": 0.6437, "T2_labels": 10, "T4": 0.5515, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.98, "T2": 28.75, "T4": 10.3, "T3": null }, "T1_bins": { "zero": 0.97, "1-10": 0.9133, "11-100": 0.9142, "101-1000": 0.9008, ">1000": 0.873 }, "S": 0.0379, "S_ci": [ 0.0351, 0.0408 ], "S_n": 4310, "T1_asfound_1-10": 0.9512 }, "all": { "index": 44.31, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9122, "T2": 0.7079, "T2_labels": 11, "T4": 0.5445, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 82.44, "T2": 41.58, "T4": 8.91, "T3": null }, "S": 0.0379, "S_ci": [ 0.0351, 0.0408 ], "S_n": 4310, "T1_asfound_1-10": 0.9512 } } } }, { "track": "F", "tag": "t23l__mpnet", "name": "all-mpnet-base-v2", "short": "all-mpnet-base-v2", "version": "v1", "in_v0": true, "modality": "text embedder", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "all-mpnet-base-v2", "text_label": "Scored on corpus L (GenePT's 2023 text, generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "MPNet fine-tuned on one billion sentence pairs", "date": "2021 (base model 2020)", "signal": "general text", "source": "https://huggingface.co/sentence-transformers/all-mpnet-base-v2", "cite": "Song et al., 2020; Reimers and Gurevych, 2019" }, "probes": { "linear": { "0-10": { "index": 33.34, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8196, "T2": 0.6258, "T2_labels": 10, "T4": 0.5548, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 63.91, "T2": 25.16, "T4": 10.95, "T3": null }, "T1_bins": { "zero": 0.9112, "1-10": 0.8075, "11-100": 0.8405, "101-1000": 0.8469, ">1000": 0.828 }, "S": 0.0884, "S_ci": [ 0.0822, 0.0944 ], "S_n": 4310, "T1_asfound_1-10": 0.8959 }, "all": { "index": 39.06, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8353, "T2": 0.6998, "T2_labels": 11, "T4": 0.5508, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 67.06, "T2": 39.95, "T4": 10.15, "T3": null }, "S": 0.0884, "S_ci": [ 0.0822, 0.0944 ], "S_n": 4310, "T1_asfound_1-10": 0.8959 } }, "mlp": { "0-10": { "index": 41.02, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9207, "T2": 0.6398, "T2_labels": 10, "T4": 0.5548, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.14, "T2": 27.96, "T4": 10.95, "T3": null }, "T1_bins": { "zero": 0.9664, "1-10": 0.9146, "11-100": 0.9181, "101-1000": 0.906, ">1000": 0.8785 }, "S": 0.0382, "S_ci": [ 0.0354, 0.0411 ], "S_n": 4310, "T1_asfound_1-10": 0.9529 }, "all": { "index": 44.95, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9155, "T2": 0.708, "T2_labels": 11, "T4": 0.5508, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.1, "T2": 41.6, "T4": 10.15, "T3": null }, "S": 0.0382, "S_ci": [ 0.0354, 0.0411 ], "S_n": 4310, "T1_asfound_1-10": 0.9529 } } } }, { "track": "F", "tag": "t23l__bge", "name": "bge-large-en-v1.5", "short": "bge-large-en-v1.5", "version": "v1", "in_v0": true, "modality": "text embedder", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "bge-large-en-v1.5", "text_label": "Scored on corpus L (GenePT's 2023 text, generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "BERT-large trained on web text pairs", "date": "Sep 2023", "signal": "general text", "source": "https://huggingface.co/BAAI/bge-large-en-v1.5", "cite": "Xiao et al., 2023" }, "probes": { "linear": { "0-10": { "index": 32.31, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8208, "T2": 0.6113, "T2_labels": 10, "T4": 0.5525, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 64.17, "T2": 22.26, "T4": 10.49, "T3": null }, "T1_bins": { "zero": 0.9202, "1-10": 0.8077, "11-100": 0.8455, "101-1000": 0.8506, ">1000": 0.8344 }, "S": 0.0921, "S_ci": [ 0.0859, 0.0983 ], "S_n": 4310, "T1_asfound_1-10": 0.8999 }, "all": { "index": 39.11, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.839, "T2": 0.697, "T2_labels": 11, "T4": 0.5506, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 67.8, "T2": 39.4, "T4": 10.12, "T3": null }, "S": 0.0921, "S_ci": [ 0.0859, 0.0983 ], "S_n": 4310, "T1_asfound_1-10": 0.8999 } }, "mlp": { "0-10": { "index": 41.0, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9196, "T2": 0.6429, "T2_labels": 10, "T4": 0.5525, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.93, "T2": 28.58, "T4": 10.49, "T3": null }, "T1_bins": { "zero": 0.9677, "1-10": 0.9133, "11-100": 0.9177, "101-1000": 0.9059, ">1000": 0.8812 }, "S": 0.0402, "S_ci": [ 0.0373, 0.0431 ], "S_n": 4310, "T1_asfound_1-10": 0.9535 }, "all": { "index": 45.51, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.915, "T2": 0.717, "T2_labels": 11, "T4": 0.5506, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 83.0, "T2": 43.41, "T4": 10.12, "T3": null }, "S": 0.0402, "S_ci": [ 0.0373, 0.0431 ], "S_n": 4310, "T1_asfound_1-10": 0.9535 } } } }, { "track": "F", "tag": "t23l__pubmedbert", "name": "pubmedbert-base-embeddings", "short": "pubmedbert-base-embeddings", "version": "v1", "in_v0": true, "modality": "text embedder", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "pubmedbert-base-embeddings", "text_label": "Scored on corpus L (GenePT's 2023 text, generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "PubMedBERT (PubMed abstracts and full text) fine-tuned for sentence embeddings", "date": "2023 (base model 2020)", "signal": "biomedical literature", "source": "https://huggingface.co/NeuML/pubmedbert-base-embeddings", "cite": "Gu et al., 2021" }, "probes": { "linear": { "0-10": { "index": 34.64, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.822, "T2": 0.6409, "T2_labels": 10, "T4": 0.5567, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 64.39, "T2": 28.18, "T4": 11.33, "T3": null }, "T1_bins": { "zero": 0.9118, "1-10": 0.8101, "11-100": 0.8522, "101-1000": 0.8656, ">1000": 0.8573 }, "S": 0.0896, "S_ci": [ 0.0831, 0.0961 ], "S_n": 4310, "T1_asfound_1-10": 0.8998 }, "all": { "index": 41.37, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8461, "T2": 0.7235, "T2_labels": 11, "T4": 0.551, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 69.23, "T2": 44.69, "T4": 10.2, "T3": null }, "S": 0.0896, "S_ci": [ 0.0831, 0.0961 ], "S_n": 4310, "T1_asfound_1-10": 0.8998 } }, "mlp": { "0-10": { "index": 41.95, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9235, "T2": 0.649, "T2_labels": 10, "T4": 0.5567, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.71, "T2": 29.8, "T4": 11.33, "T3": null }, "T1_bins": { "zero": 0.9707, "1-10": 0.9173, "11-100": 0.9237, "101-1000": 0.9153, ">1000": 0.8941 }, "S": 0.0404, "S_ci": [ 0.0376, 0.0433 ], "S_n": 4310, "T1_asfound_1-10": 0.9578 }, "all": { "index": 46.19, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9213, "T2": 0.7207, "T2_labels": 11, "T4": 0.551, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.25, "T2": 44.13, "T4": 10.2, "T3": null }, "S": 0.0404, "S_ci": [ 0.0376, 0.0433 ], "S_n": 4310, "T1_asfound_1-10": 0.9578 } } } }, { "track": "F", "tag": "t23l__tfidf", "name": "TF-IDF (1,024-d SVD)", "short": "TF-IDF (1,024-d SVD)", "version": "v1", "in_v0": true, "modality": "text embedder", "cls": "clean", "text": "leakfree", "train_text": null, "encoder": "TF-IDF (1,024-d SVD)", "text_label": "Scored on corpus L (GenePT's 2023 text, generated summaries removed)", "gen_share": 0.0, "n_genes": 19256, "meta": { "training": "fitted on the two corpora themselves (1,024-d SVD)", "date": "—", "signal": "none", "source": "https://github.com/hwang595/gene-embedding-exposure", "cite": "this work" }, "probes": { "linear": { "0-10": { "index": 32.47, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.83, "T2": 0.6125, "T2_labels": 10, "T4": 0.5445, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 66.01, "T2": 22.5, "T4": 8.9, "T3": null }, "T1_bins": { "zero": 0.9027, "1-10": 0.8204, "11-100": 0.8162, "101-1000": 0.8107, ">1000": 0.7946 }, "S": 0.0429, "S_ci": [ 0.0382, 0.0483 ], "S_n": 4310, "T1_asfound_1-10": 0.8634 }, "all": { "index": 34.99, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.8186, "T2": 0.674, "T2_labels": 11, "T4": 0.5323, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 63.72, "T2": 34.81, "T4": 6.45, "T3": null }, "S": 0.0429, "S_ci": [ 0.0382, 0.0483 ], "S_n": 4310, "T1_asfound_1-10": 0.8634 } }, "mlp": { "0-10": { "index": 39.33, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9202, "T2": 0.6252, "T2_labels": 10, "T4": 0.5445, "T1_n": 4878, "T4_n": 6077, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 84.04, "T2": 25.04, "T4": 8.9, "T3": null }, "T1_bins": { "zero": 0.9594, "1-10": 0.9151, "11-100": 0.9017, "101-1000": 0.8782, ">1000": 0.8456 }, "S": 0.021, "S_ci": [ 0.0188, 0.0233 ], "S_n": 4310, "T1_asfound_1-10": 0.936 }, "all": { "index": 39.44, "ci": null, "rank": null, "rank_ci": null, "index4": null, "T1": 0.9011, "T2": 0.6583, "T2_labels": 11, "T4": 0.5323, "T1_n": null, "T4_n": null, "T3": null, "T3_n": null, "T3_novel": null, "skills": { "T1": 80.22, "T2": 31.66, "T4": 6.45, "T3": null }, "S": 0.021, "S_ci": [ 0.0188, 0.0233 ], "S_n": 4310, "T1_asfound_1-10": 0.936 } } } } ], "sources": { "round4/results/r11/bench_v1.json": "909d84c92617e431cd201cec037d0c4735e8167e383602a00b9f50f6eb0c15eb", "round4/results/r14/rows_v11.json": "c2ffb367be03d8ece88ff4626b19310a4e2d163ad84a787a0e24351409574cbc", "round4/results/r14/bench_r14.json": "7fa5c54007ab6aed811a64c7ca091d5b0ddd24b3ab7e5a5a64a11535af5dc3a9", "round4/results/r14/t4_recall.json": "6efd43f51b7ea56b116048876c613a3cf47349bae528cb1b0256496861216751", "round4/results/r14/coverage.json": "fcc82c2131e1d8224b4a81b3eda5835b5dd6d68ab031037e4ccbcde068efd280", "round4/release/provenance_human_protein_coding.tsv": "f38c62cf601948176c88f5382ee1043d0e2d00f73c3d48a4f8b2ae007faeb6a6" } }