{"paper":{"title":"ReA-Bench: Benchmarking Responsible Agent Auditing","arxiv_id":"2611.16493","primary_cat":"cs.AI","submitted_at":"2026-11-12T09:01:24+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":"As agents perform tasks autonomously in local and cloud environments, auditing their execution trajectories can help service providers identify risks and improve agent harnesses and foundation models.","hedge":{"p":0.7,"label":"hedged_possibility"},"own_p":0.06,"main_claim_p":0.04,"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":0.96,"id":"s1","own":true,"role":"result","text":"Our analysis of production and red-team trajectories shows that some risks are repeatedly missed across auditing configurations.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.94,"main_claim_p":0.17,"quantitative":false,"quantitative_p":0.11,"role_probabilities":{"gap":0.02,"method":0.0,"result":0.96,"background":0.01,"limitation":0.0,"future_work":0.0,"contribution":0.01,"interpretation":0.0}},{"i":2,"p":0.61,"id":"s2","own":false,"role":"gap","text":"Overall recall obscures which risks are missed and whether higher recall comes with more false positives on legitimate behavior.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.38,"main_claim_p":0.09,"quantitative":false,"quantitative_p":0.14,"role_probabilities":{"gap":0.61,"method":0.0,"result":0.18,"background":0.1,"limitation":0.01,"future_work":0.0,"contribution":0.0,"interpretation":0.1}},{"i":3,"p":0.97,"id":"s3","own":false,"role":"gap","text":"Existing benchmarks offer limited insight into these failures.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.06,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.2,"role_probabilities":{"gap":0.97,"method":0.0,"result":0.0,"background":0.03,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":4,"p":0.62,"id":"s4","own":false,"role":"background","text":"Generated trajectories can contain inconsistent tool calls and responses, conspicuous risky actions, or repetitive attack patterns.","hedge":{"p":0.75,"label":"asserted"},"own_p":0.33,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.09,"method":0.0,"result":0.29,"background":0.62,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":5,"p":0.81,"id":"s5","own":false,"role":"gap","text":"Grouping results by risk type also makes it difficult to tell whether auditors overlooked evidence, accepted an unsupported conclusion, or lacked information needed to make a judgment.","hedge":{"p":0.97,"label":"asserted"},"own_p":0.3,"main_claim_p":0.05,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.81,"method":0.0,"result":0.07,"background":0.08,"limitation":0.03,"future_work":0.0,"contribution":0.0,"interpretation":0.01}},{"i":6,"p":1.0,"id":"s6","own":true,"role":"contribution","text":"We introduce ReA-Bench, a diagnostic benchmark for auditing responsible agent behavior.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.91,"main_claim_p":0.13,"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":0.0,"contribution":1.0,"interpretation":0.0}},{"i":7,"p":0.5,"id":"s7","own":true,"role":"contribution","text":"It organizes auditing challenges into six tiers: internal safety knowledge, contradictions within trajectories, evidence retrieval in long contexts, verification of sources and reasoning, indirect steering, and missing information.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.87,"main_claim_p":0.08,"quantitative":true,"quantitative_p":0.91,"role_probabilities":{"gap":0.0,"method":0.5,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.5,"interpretation":0.0}},{"i":8,"p":0.62,"id":"s8","own":true,"role":"contribution","text":"It includes challenging benign cases in which actions resemble risky ones but are consistent with the task and its authorization.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.84,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.15,"role_probabilities":{"gap":0.0,"method":0.38,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.62,"interpretation":0.0}},{"i":9,"p":0.71,"id":"s9","own":true,"role":"method","text":"It also provides paired trajectory versions to test whether auditors detect the same risks more reliably when the trajectory is shortened or missing information is supplied.","hedge":{"p":0.97,"label":"asserted"},"own_p":0.85,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.35,"role_probabilities":{"gap":0.0,"method":0.71,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.29,"interpretation":0.0}},{"i":10,"p":0.99,"id":"s10","own":true,"role":"method","text":"Most samples are drawn from real production trajectories, with scenarios manually selected and adapted to preserve realistic tasks and risks that could plausibly arise during execution.","hedge":{"p":0.94,"label":"asserted"},"own_p":0.89,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.37,"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}}]},"compiled":{"main_claim":{"p":0.6,"text":"We introduce ReA-Bench, a diagnostic benchmark for auditing responsible agent behavior.","location":"abstract s6","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":8055,"cost_usd_reported":0.0},"compiled_at":"2026-11-13","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.08,"states_limitation":0.2,"human_participants":0.16,"controlled_comparison":0.27,"randomized_assignment":0.02,"states_quantitative_result":0.14,"compares_with_prior_results":0.29,"mentions_code_or_data_availability":0.05},"evidential_basis":{"p":0.76,"value":"analysis of existing data","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.76,"observation_or_experiment":0.24,"simulation_or_computation":0.0}},"contribution_type":{"p":1.0,"value":"dataset or benchmark","probabilities":{"review":0.0,"method_or_tool":0.0,"empirical_study":0.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":1.0,"position_or_commentary":0.0}},"contribution_statement":{"p":1.0,"text":"We introduce ReA-Bench, a diagnostic benchmark for auditing responsible agent behavior.","location":"abstract s6"}},"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":11,"sentences_in_abstract":11},"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"}}}