{"paper":{"title":"HarnessEvolve: Learning from Reference Trajectories for Reliable Agent Self-Evolution","arxiv_id":"2609.00829","primary_cat":"cs.LG","submitted_at":"2026-09-01T07:31:18+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":"Self-evolving agents advance toward autonomy by optimizing their harness---prompts, skills, tools, and execution logic---based on environmental feedback.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.06,"main_claim_p":0.03,"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":0.68,"id":"s1","own":false,"role":"background","text":"This paradigm, however, is hampered by three challenges: \\textit{credit assignment failure}, where terminal success/failure feedback makes it ambiguous which step caused the error; \\textit{shortcut learning}, where agents memorize task-specific patterns rather than acquire generalizable capabilities; and \\textit{catastrophic forgetting}, where unguarded updates degrade previously acquired competence.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.07,"main_claim_p":0.02,"quantitative":true,"quantitative_p":0.7,"role_probabilities":{"gap":0.32,"method":0.0,"result":0.0,"background":0.68,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":1.0,"id":"s2","own":true,"role":"contribution","text":"In this paper, we introduce HarnessEvolve, a self-evolving framework that learns from reference trajectories to achieve reliable agent self-evolution.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.81,"main_claim_p":0.1,"quantitative":false,"quantitative_p":0.03,"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":3,"p":0.81,"id":"s3","own":true,"role":"method","text":"HarnessEvolve decouples the execution agent from the evolutionary pipeline, assigning execution, evaluation, optimization, and gating to independent agent modules, enabling generalizable and stable harness improvements.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.85,"main_claim_p":0.1,"quantitative":false,"quantitative_p":0.06,"role_probabilities":{"gap":0.0,"method":0.81,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.19,"interpretation":0.0}},{"i":4,"p":0.98,"id":"s4","own":true,"role":"method","text":"Specifically, HarnessEvolve overcomes credit assignment failure by generating reference trajectories (execution paths produced when given the ground-truth answers) and aligning failed executions against them to extract error signals, which are clustered to reveal systematic failure patterns.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.89,"main_claim_p":0.04,"quantitative":false,"quantitative_p":0.06,"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":"To prevent shortcut learning and catastrophic forgetting, candidate harness updates must pass two gates: a quality gate that filters data leakage and prompt bloat, and a performance gate that accepts each update if it improves on the current batch without degrading recent batches, with epoch-end validation on a held-out set selecting the best-performing accepted agent snapshot.","hedge":{"p":0.99,"label":"asserted"},"own_p":0.81,"main_claim_p":0.05,"quantitative":true,"quantitative_p":0.75,"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":0.95,"id":"s6","own":true,"role":"method","text":"We conduct extensive experiments on several benchmarks spanning open-domain and enterprise scenarios, using different models and agent frameworks.","hedge":{"p":0.91,"label":"asserted"},"own_p":0.95,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.05,"role_probabilities":{"gap":0.0,"method":0.95,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.05,"interpretation":0.0}},{"i":7,"p":0.81,"id":"s7","own":true,"role":"result","text":"Results demonstrate that HarnessEvolve consistently outperforms state-of-the-art baselines across all benchmarks and settings, confirming reliability across task domains.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.96,"main_claim_p":0.86,"quantitative":true,"quantitative_p":0.81,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.81,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.19}}]},"compiled":{"main_claim":{"p":1.0,"text":"Results demonstrate that HarnessEvolve consistently outperforms state-of-the-art baselines across all benchmarks and settings, confirming reliability across task domains.","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":6519,"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":0.05,"states_limitation":0.09,"human_participants":0.03,"controlled_comparison":0.9,"randomized_assignment":0.02,"states_quantitative_result":0.05,"compares_with_prior_results":0.97,"mentions_code_or_data_availability":0.03},"evidential_basis":{"p":1.0,"value":"observation or experiment","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.0,"observation_or_experiment":1.0,"simulation_or_computation":0.0}},"contribution_type":{"p":1.0,"value":"method or tool","probabilities":{"review":0.0,"method_or_tool":1.0,"empirical_study":0.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.99,"text":"In this paper, we introduce HarnessEvolve, a self-evolving framework that learns from reference trajectories to achieve reliable agent self-evolution.","location":"abstract s2"}},"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":8,"sentences_in_abstract":8},"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"}}}