{"paper":{"title":"Auditing the Product, Not the Recursion: The Evaluation Evidence Behind Recursive Self-Improvement Claims","arxiv_id":"2610.27610","primary_cat":"cs.LG","submitted_at":"2026-10-20T04:23:05+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":"Recursive self-improvement (RSI) promises AI systems, such as self-evolving LLM agents, that become better at producing further improvements.","hedge":{"p":0.78,"label":"asserted"},"own_p":0.04,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.03,"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":0.91,"id":"s1","own":false,"role":"background","text":"Evaluating this promise requires separating product claims about the final system from process claims about the improvement process itself.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.07,"main_claim_p":0.05,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.07,"method":0.0,"result":0.0,"background":0.91,"limitation":0.0,"future_work":0.0,"contribution":0.01,"interpretation":0.01}},{"i":2,"p":0.96,"id":"s2","own":false,"role":"gap","text":"Existing surveys specify the evidence these claims require, but how often studies report that evidence remains unmeasured.","hedge":{"p":0.79,"label":"unclear"},"own_p":0.06,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.96,"method":0.0,"result":0.0,"background":0.04,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":3,"p":0.92,"id":"s3","own":true,"role":"method","text":"We examine this question through full-text coding of 114 eligible studies in a reproducible arXiv search frame frozen in October 2026, including 90 that describe their own systems as RSI.","hedge":{"p":0.97,"label":"asserted"},"own_p":0.95,"main_claim_p":0.03,"quantitative":true,"quantitative_p":0.98,"role_probabilities":{"gap":0.0,"method":0.92,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.08,"interpretation":0.0}},{"i":4,"p":0.97,"id":"s4","own":true,"role":"method","text":"The codebook links claims to evaluation designs.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.78,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.03,"role_probabilities":{"gap":0.0,"method":0.97,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.03,"interpretation":0.0}},{"i":5,"p":0.99,"id":"s5","own":true,"role":"method","text":"Independent human coding, third-reader adjudication, and a separate blind coding of improver comparisons assess coding reliability.","hedge":{"p":0.87,"label":"asserted"},"own_p":0.88,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.05,"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":1.0,"id":"s6","own":true,"role":"result","text":"Among the 90 RSI studies, 90% report product evaluation beyond development tasks under a broad criterion that includes partial or unstated separation, while 60% report evaluation on tasks unused for selection.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.92,"main_claim_p":0.12,"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":7,"p":1.0,"id":"s7","own":true,"role":"result","text":"Only 22% audit two or more reruns of the full loop.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.87,"main_claim_p":0.08,"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":1.0,"id":"s8","own":true,"role":"result","text":"In a human-coded random sample, the RSI studies show the same contrast: 45% evaluate on tasks unused for selection and 14% audit two or more reruns.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.91,"main_claim_p":0.07,"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":9,"p":1.0,"id":"s9","own":true,"role":"result","text":"Of the 26 studies claiming a better improvement procedure, 23 report a related comparison, but only 4 compare earlier and later procedures from identical starting artifacts at a matched budget.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.87,"main_claim_p":0.06,"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":10,"p":0.51,"id":"s10","own":true,"role":"interpretation","text":"Across these results and a comparison set of established works, evidence for repeatability and improved procedures remains less developed than product evaluation.","hedge":{"p":0.93,"label":"asserted"},"own_p":0.82,"main_claim_p":0.46,"quantitative":true,"quantitative_p":0.87,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.49,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.51}},{"i":11,"p":0.89,"id":"s11","own":true,"role":"interpretation","text":"Three propositions explain how selection effects, shared evaluator errors, and changes in system state limit what product evaluations can establish.","hedge":{"p":0.48,"label":"asserted"},"own_p":0.6,"main_claim_p":0.09,"quantitative":false,"quantitative_p":0.46,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.04,"future_work":0.0,"contribution":0.07,"interpretation":0.89}},{"i":12,"p":0.58,"id":"s12","own":true,"role":"interpretation","text":"Together, the findings support a claim-indexed reporting standard that connects each claim to the evaluation design needed to test it.","hedge":{"p":0.73,"label":"asserted"},"own_p":0.51,"main_claim_p":0.75,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.39,"contribution":0.03,"interpretation":0.58}}]},"compiled":{"main_claim":{"p":0.56,"text":"Across these results and a comparison set of established works, evidence for repeatability and improved procedures remains less developed than product evaluation.","location":"abstract s10","hedge_strength":{"p":0.93,"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":9513,"cost_usd_reported":0.0},"compiled_at":"2026-10-21","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.05,"states_limitation":0.29,"human_participants":0.5,"controlled_comparison":0.38,"randomized_assignment":0.04,"states_quantitative_result":0.99,"compares_with_prior_results":0.81,"mentions_code_or_data_availability":0.11},"evidential_basis":{"p":0.65,"value":"analysis of existing data","probabilities":{"argument":0.0,"literature":0.33,"proof_or_derivation":0.0,"analysis_of_existing_data":0.65,"observation_or_experiment":0.02,"simulation_or_computation":0.0}},"contribution_type":{"p":0.58,"value":"empirical study","probabilities":{"review":0.42,"method_or_tool":0.0,"empirical_study":0.58,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":1.0,"text":"We examine this question through full-text coding of 114 eligible studies in a reproducible arXiv search frame frozen in October 2026, including 90 that describe their own systems as RSI.","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":13,"sentences_in_abstract":13},"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"}}}