{"paper":{"title":"Continue or Restart? The Limits of Early Failure Prediction in Coding Agents","arxiv_id":"2610.35971","primary_cat":"cs.SE","submitted_at":"2026-10-26T04:50:03+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":"Early stopping promises to reduce compute costs for autonomous coding agents by terminating unpromising runs early.","hedge":{"p":0.55,"label":"hedged_possibility"},"own_p":0.05,"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":0.71,"id":"s1","own":false,"role":"background","text":"Deciding when to abandon a run requires separating baseline issue difficulty from the trajectory of an individual attempt.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.07,"main_claim_p":0.03,"quantitative":false,"quantitative_p":0.04,"role_probabilities":{"gap":0.29,"method":0.0,"result":0.0,"background":0.71,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.0}},{"i":2,"p":0.85,"id":"s2","own":true,"role":"method","text":"We study the distinction in 147,002 execution traces from two public corpora spanning four model-framework settings.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.95,"main_claim_p":0.02,"quantitative":true,"quantitative_p":0.97,"role_probabilities":{"gap":0.0,"method":0.85,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.15,"interpretation":0.0}},{"i":3,"p":0.99,"id":"s3","own":true,"role":"method","text":"Comparing pre-execution issue context against early execution prefixes, we evaluate linear count models, initial-turn text embeddings, task-conditional rankers, and a fine-tuned Qwen3-0.6B monitor.","hedge":{"p":0.9,"label":"asserted"},"own_p":0.93,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.47,"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":4,"p":1.0,"id":"s4","own":true,"role":"result","text":"Although early prefixes modestly improve cross-task outcome ranking, within-task discrimination remains near chance across early landmarks (5k-20k tokens): initial-turn embeddings and 20k-token task-conditional models yield AUCs around 0.50, and three-seed neural evaluations across 32,406 attempts at 20k tokens achieve a mean within-task AUC of 0.50 [0.48, 0.51] across tasks of varying solve rates.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.96,"main_claim_p":0.23,"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":5,"p":0.98,"id":"s5","own":true,"role":"method","text":"To evaluate how these predictions translate into stopping decisions, we replay score-guided and matched random restarts under shared token budgets alongside uninterrupted continuation.","hedge":{"p":0.7,"label":"asserted"},"own_p":0.93,"main_claim_p":0.02,"quantitative":false,"quantitative_p":0.19,"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":6,"p":1.0,"id":"s6","own":true,"role":"result","text":"Under a 75k-token cap, learned restarting improves resolution over matched random restarting by up to 1.98 percentage points.","hedge":{"p":0.94,"label":"asserted"},"own_p":0.95,"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":7,"p":0.64,"id":"s7","own":true,"role":"interpretation","text":"An exact algebraic decomposition resolves this paradox: the gain over random restarting stems almost entirely from between-task stopping allocation (avoiding stops on easy issues), whereas attempt-level selection within a task contributes near zero (-0.11 to +0.24 percentage points).","hedge":{"p":1.0,"label":"asserted"},"own_p":0.95,"main_claim_p":0.21,"quantitative":true,"quantitative_p":0.99,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.36,"background":0.0,"limitation":0.0,"future_work":0.0,"contribution":0.0,"interpretation":0.64}},{"i":8,"p":1.0,"id":"s8","own":true,"role":"result","text":"Across all four settings at the 75k cap (and 216 of 224 tested budget-threshold configurations overall), uninterrupted continuation achieves higher solve rates at lower mean token cost than both restart policies.","hedge":{"p":1.0,"label":"asserted"},"own_p":0.95,"main_claim_p":0.62,"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":0.63,"id":"s9","own":false,"role":"interpretation","text":"Practical evaluations of early stopping must therefore establish attempt-level discrimination and demonstrate net compute savings relative to continuation.","hedge":{"p":0.98,"label":"asserted"},"own_p":0.33,"main_claim_p":0.76,"quantitative":false,"quantitative_p":0.12,"role_probabilities":{"gap":0.0,"method":0.0,"result":0.0,"background":0.0,"limitation":0.0,"future_work":0.37,"contribution":0.0,"interpretation":0.63}}]},"compiled":{"main_claim":{"p":0.85,"text":"Across all four settings at the 75k cap (and 216 of 224 tested budget-threshold configurations overall), uninterrupted continuation achieves higher solve rates at lower mean token cost than both restart policies.","location":"abstract s8","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":7793,"cost_usd_reported":0.0},"compiled_at":"2026-10-28","bank_version":"abstract-p1-20260923","model_version":"classifier-1.13.0","abstract_states":{"claims_first":0.04,"states_limitation":0.87,"human_participants":0.03,"controlled_comparison":0.97,"randomized_assignment":0.08,"states_quantitative_result":0.99,"compares_with_prior_results":0.95,"mentions_code_or_data_availability":0.03},"evidential_basis":{"p":0.67,"value":"observation or experiment","probabilities":{"argument":0.0,"literature":0.0,"proof_or_derivation":0.0,"analysis_of_existing_data":0.29,"observation_or_experiment":0.67,"simulation_or_computation":0.04}},"contribution_type":{"p":1.0,"value":"empirical study","probabilities":{"review":0.0,"method_or_tool":0.0,"empirical_study":1.0,"simulation_study":0.0,"theoretical_result":0.0,"dataset_or_benchmark":0.0,"position_or_commentary":0.0}},"contribution_statement":{"p":0.54,"text":"We study the distinction in 147,002 execution traces from two public corpora spanning four model-framework settings.","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":10,"sentences_in_abstract":10},"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"}}}