Newest papers

Enhancing Bug-Exposing Test Generation via Step-Level Credit Assignment and Self-Distillation from Valuable Failures

Wenjie Gan, Zhaoxi Zhang, Jiahui Liang, Weikang Li, Wanyuan Wang, Li Zhang, Jian Dong · 2026 · arXiv

DOI: 10.48550/arXiv.2610.27694 · open-access PDF

French

Les grands modèles de langues ont montré de fortes capacités de génération de code, mais leurs sorties contiennent souvent des bugs subtils. La réparation automatique de ce type de code nécessite des tests informatifs: des entrées qui exposent un comportement défectueux, associées à des sorties correctes attendues. Les méthodes récentes de génération de test basées sur RL appliquent la même récompense de résultat à toutes les étapes de raisonnement, ce qui peut décourager le raisonnement utile et renforcer les étapes défectueuses.

Machine translation (Meta M2M100); the English original follows.

Abstract

Large Language Models have shown strong code generation capabilities, yet their outputs often contain subtle bugs. Automatically repairing such code requires informative tests: inputs that expose faulty behavior, paired with correct expected outputs. Recent RL-based test generation methods apply the same outcome reward to all reasoning steps, which can discourage useful reasoning and reinforce flawed steps. They also fail to effectively learn from trajectories that find bug-triggering inputs but predict incorrect outputs, leaving their verified bug-triggering progress insufficiently exploited.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.27694, 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.txt

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange