{"paper":{"title":"Beyond Gain: Measuring Skill Internalization for Agent Skill Retirement","arxiv_id":"2610.32322","primary_cat":"cs.AI","submitted_at":"2026-10-23T08:21:22+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":"LLM agents retrieve skills from public hubs that hold up to about 100,000 entries and serve many models.","hedge":{"p":0.97,"label":"asserted"},"own_p":0.08,"main_claim_p":0.02,"quantitative":true,"quantitative_p":0.97,"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":0.85,"id":"s1","own":false,"role":"gap","text":"As models improve, a hub must find the skills a given model has internalized, meaning it can do without them, and retire them for that model, yet internalization has not been measured with known reliability.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.08,"main_claim_p":0.08,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.85,"method":0.0,"result":0.0,"background":0.15,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":0.85,"id":"s2","own":false,"role":"background","text":"Published rules that compare runs with and without a skill retire on the gain alone.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.05,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.31,"role_probabilities":{"gap":0.15,"method":0.0,"result":0.0,"background":0.85,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":3,"p":0.45,"id":"s3","own":true,"role":"result","text":"The gain is near zero both for an internalized skill and when neither the model nor the skill delivers the capability, a state we call uncovered.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.51,"main_claim_p":0.12,"quantitative":true,"quantitative_p":0.75,"role_probabilities":{"gap":0.14,"method":0.0,"result":0.45,"background":0.33,"limitation":0.01,"future_work":0.0,"contribution":0.0,"interpretation":0.07}},{"i":4,"p":0.9,"id":"s4","own":true,"role":"method","text":"We instead withhold the skill, score the model on a task synthesized from it, and trust a high unaided score only when another model gains from the skill, since nearly half of high scores come from tasks no model benefits from.","hedge":{"p":0.97,"label":"asserted"},"own_p":0.86,"main_claim_p":0.38,"quantitative":true,"quantitative_p":0.96,"role_probabilities":{"gap":0.0,"method":0.9,"result":0.01,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.09,"interpretation":0.0}},{"i":5,"p":1.0,"id":"s5","own":true,"role":"result","text":"Under four models, a published gain rule applied with one task per skill removes over half of the uncovered skills.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.69,"main_claim_p":0.16,"quantitative":true,"quantitative_p":0.98,"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":6,"p":1.0,"id":"s6","own":true,"role":"result","text":"On SkillsBench, 28 of its 31 retirements rest on probes no run passed in full, while retiring on the unaided score costs no detectable reward.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.85,"main_claim_p":0.19,"quantitative":true,"quantitative_p":0.98,"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":7,"p":0.9,"id":"s7","own":true,"role":"result","text":"Unaided verdicts reproduce across tasks more often than the gain rule's (Cohen's kappa 0.42 against 0.23) but differ across models, so a hub must read them per model.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.91,"main_claim_p":0.18,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.9,"background":0.0,"limitation":0.01,"future_work":0.0,"contribution":0.0,"interpretation":0.09}},{"i":8,"p":0.95,"id":"s8","own":true,"role":"contribution","text":"We release 1,251 skills with tasks and item scores under four models.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.94,"main_claim_p":0.04,"quantitative":true,"quantitative_p":0.97,"role_probabilities":{"gap":0.0,"method":0.05,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.95,"interpretation":0.0}}]},"compiled":{"main_claim":{"p":0.8,"text":"We instead withhold the skill, score the model on a task synthesized from it, and trust a high unaided score only when another model gains from the skill, since nearly half of high scores come from tasks no model benefits from.","location":"abstract s4","hedge_strength":{"p":0.97,"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":6972,"cost_usd_reported":0.0},"compiled_at":"2026-10-26","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.1,"states_limitation":0.91,"human_participants":0.05,"controlled_comparison":0.79,"randomized_assignment":0.02,"states_quantitative_result":0.99,"compares_with_prior_results":0.95,"mentions_code_or_data_availability":0.84},"evidential_basis":{"p":0.97,"value":"observation or experiment","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.01,"observation_or_experiment":0.97,"simulation_or_computation":0.02}},"contribution_type":{"p":0.56,"value":"dataset or benchmark","probabilities":{"review":0.0,"method_or_tool":0.3,"empirical_study":0.14,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.56,"position_or_commentary":0.0}},"contribution_statement":{"p":0.97,"text":"We instead withhold the skill, score the model on a task synthesized from it, and trust a high unaided score only when another model gains from the skill, since nearly half of high scores come from tasks no model benefits from.","location":"abstract s4"}},"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,"review_comments_judged":{"judged_by":["gpt-5.6-luna","grok-4.7"],"judged_against":"Pith's own text copy of each paper: text extracted from the PDF, the first 90,000 characters","measured_on":"2026-10-04","accurate":113,"decided":149,"sampled_reviews":200,"split_by_lane":false,"by_lane":"https://pith.science/accuracy#by-lane"}}}