{"paper":{"title":"Efficient Auto-Interpretability of AI Models in Biology","arxiv_id":"2608.27754","primary_cat":"q-bio.QM","submitted_at":"2026-08-27T22:32:43+00:00","version_compiled":"v1"},"notice":"Every quoted sentence is the authors' text from the arXiv listing abstract. Labels and probabilities are the outputs of classifier-1.13.0 on the question bank abstract-p1-20260923, exactly as returned. No sentence in this record was written by a language model.","schema":"pith.compiled-record/v0","status":"compiled","abstract":{"source":"arXiv listing abstract v1","sentences":[{"i":0,"p":1.0,"id":"s0","own":false,"role":"background","text":"Sparse autoencoders (SAEs), and other interpretability methods could turn AI models in Biology and other fields into engines of scientific discovery by explaining the superhuman capabilities of those models.","hedge":{"p":1.0,"label":"hedged_possibility"},"own_p":0.05,"main_claim_p":0.08,"quantitative":false,"quantitative_p":0.19,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":1.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":1,"p":0.63,"id":"s1","own":false,"role":"background","text":"However, a latent is only useful if we know three things: whether it is coherent, whether it can be described, and whether that description has predictive power.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.08,"main_claim_p":0.07,"quantitative":true,"quantitative_p":0.66,"role_probabilities":{"gap":0.37,"method":0.0,"result":0.0,"background":0.63,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":0.53,"id":"s2","own":false,"role":"gap","text":"These questions are routinely conflated.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.11,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.53,"method":0.0,"result":0.0,"background":0.47,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":3,"p":0.92,"id":"s3","own":true,"role":"contribution","text":"We assemble them into a single pipeline and report the practical innovations each stage required.","hedge":{"p":0.97,"label":"asserted"},"own_p":0.9,"main_claim_p":0.07,"quantitative":false,"quantitative_p":0.03,"role_probabilities":{"gap":0.0,"method":0.08,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.92,"interpretation":0.0}},{"i":4,"p":0.98,"id":"s4","own":true,"role":"method","text":"First, cross-seed dictionary stability prioritises which latents are worth spending resources to investigate.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.79,"main_claim_p":0.06,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.0,"method":0.98,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.02,"interpretation":0.0}},{"i":5,"p":0.99,"id":"s5","own":true,"role":"method","text":"Second, an intruder-detection task asks whether a latents activating examples share a recognizable pattern.","hedge":{"p":0.93,"label":"asserted"},"own_p":0.74,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.99,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.01,"interpretation":0.0}},{"i":6,"p":0.99,"id":"s6","own":true,"role":"method","text":"Third, a separate pass proposes a candidate biological description which we convert into falsifiable predictions which can be tested in silico.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.9,"main_claim_p":0.07,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.99,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.01,"interpretation":0.0}},{"i":7,"p":1.0,"id":"s7","own":true,"role":"result","text":"Deployed on the Boltz-1 Pairformer trunk, stability prioritisation finds interpretable latents using about 4.4 times fewer latent evaluations each, and at 5.2 times lower measured cost, while recovering over half of them, and the external check shows the surfaced motifs are significantly enriched for their claimed annotations.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.96,"main_claim_p":0.42,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":1.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":8,"p":0.89,"id":"s8","own":true,"role":"interpretation","text":"The results also suggest a possible tension: the cross- seed stability might be selecting for some types of features, like structure-related ones, much more than others, such as function-related features.","hedge":{"p":0.99,"label":"hedged_possibility"},"own_p":0.76,"main_claim_p":0.14,"quantitative":true,"quantitative_p":0.52,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.03,"background":0.0,"limitation":0.08,"future_work":0.0,"contribution":0.0,"interpretation":0.89}}]},"compiled":{"main_claim":{"p":0.97,"text":"Deployed on the Boltz-1 Pairformer trunk, stability prioritisation finds interpretable latents using about 4.4 times fewer latent evaluations each, and at 5.2 times lower measured cost, while recovering over half of them, and the external check shows the surfaced motifs are significantly enriched for their claimed annotations.","location":"abstract s7","hedge_strength":{"p":1.0,"label":"asserted"}},"provenance":{"route":{"via":"TypeSafe API","model":"classifier-1.13.0","endpoint":"https://api.typesafe.ai/v1/systemone","provider":"TypeSafe"},"packing":"one request per paper","requests":1,"input_tokens":6906,"cost_usd_reported":0.0},"compiled_at":"2026-09-23","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.06,"states_limitation":0.94,"human_participants":0.04,"controlled_comparison":0.27,"randomized_assignment":0.02,"states_quantitative_result":0.98,"compares_with_prior_results":0.51,"mentions_code_or_data_availability":0.04},"evidential_basis":{"p":0.77,"value":"simulation or computation","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.0,"observation_or_experiment":0.23,"simulation_or_computation":0.77}},"contribution_type":{"p":0.99,"value":"method or tool","probabilities":{"review":0.0,"method_or_tool":0.99,"empirical_study":0.01,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.99,"text":"We assemble them into a single pipeline and report the practical innovations each stage required.","location":"abstract s3"}},"coverage":{"tier":"abstract","levels_run":["P abstract"],"not_examined":["full text","figures and tables","references and citing sentences","relations to other papers"],"sources_read":["arXiv listing abstract v1"],"levels_not_run":["S sections","T body sentences","C citations","R cross-paper relations"],"sentences_examined":9,"sentences_in_abstract":9},"measured_accuracy":{"url":"https://pith.science/accuracy","measured_on":"2026-10-02","method":"two independent model families labelled a fresh random sample blind; the figure is how often the published label matches their consensus","sentence_role_agreement":0.92,"main_claim_agreement":0.908,"typed_citation_agreement":0.933,"typed_citation_agreement_95_interval":[0.917,0.945],"review_comments_accurate":0.758,"review_comments_accurate_95_interval":[0.684,0.82],"review_caught_admitted_error_found_or_related":0.448}}