Saiyue Lyu, Jerrold Huang, Yunbei Zhang, Janet Wang, Zhenting Wang, Yingqiang Ge, Haibo Zhang · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28834 · open-access PDF
L'échelle de temps d'inference est souvent traitée comme un problème budgétaire de jeton, mais un agent de codage échoué peut rechercher des ressources qualitativement différentes: calcul, information ou rétroaction. Nous introduisons Banzhaf Resource Attribution (BRA), un cadre factoriel contrôlé qui mesure la contribution marginale de chaque ressource à la réparation et ses interactions tout en gardant l'état échoué et le protocole de révision fixe. Nous évaluons six modèles sur 194 défaillances au niveau de la fonction de MBPP+ et HumanEval+ et 126 défaillances au niveau du dépôt de SWE-bench Verified.
Machine translation (Meta M2M100); the English original follows.
Inference-time scaling is often treated as a token-budget problem, but a failed coding agent can seek qualitatively different resources: computation, information, or feedback. We introduce Banzhaf Resource Attribution (BRA), a controlled factorial framework that measures each resource's marginal contribution to repair and its interactions while holding the failed state and revision protocol fixed. We evaluate six models on 194 function-level failures from MBPP+ and HumanEval+ and 126 repository-level failures from SWE-bench Verified. Task-relevant specifications consistently improve repair, wh
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28834, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange