{"slug": "lapo-leave-one-turn-attribution-for-self-generated-process-rewards-in-multi-turn", "title": "LAPO: Leave-One-Turn Attribution for Self-Generated Process Rewards in Multi-Turn Search Reasoning", "summary": "Researchers propose LAPO, a self-generated process-supervision method for multi-turn search reasoning that uses backward leave-one-turn attribution to estimate each turn's contribution without additional reward models. LAPO achieves an average exact-match score of 0.326 across seven knowledge-intensive QA datasets, outperforming the strongest baseline by 0.053.", "body_md": "arXiv:2607.13501v1 Announce Type: new\nAbstract: Reinforcement learning for multi-turn search reasoning typically relies on terminal outcome rewards, which cannot distinguish useful, redundant, and harmful intermediate interactions. We propose LAPO, a self-generated process-supervision method based on backward leave-one-turn attribution. For each search turn, LAPO replaces the turn and its retrieval observation with a fixed [DELETE] placeholder and measures the resulting change in the current policy's mean log-likelihood of the gold answer. This Answer-Likelihood Gain estimates the turn's contribution while preserving all downstream interactions, allowing early evidence to be evaluated in the complete reasoning context. LAPO further applies sign-consistency gating, retaining only normalized process advantages whose directions agree with their raw attribution scores. The method requires no additional reward model, teacher, verifier, or LLM-as-a-Judge. Across seven knowledge-intensive question-answering datasets with local retrieval, LAPO achieves an average exact-match score of 0.326, outperforming the strongest step-reward baseline, IGPO, by 0.053. Ablations show complementary benefits from backward attribution and sign-consistency gating, demonstrating that policy-derived retrospective attribution can provide effective process supervision for multi-turn search agents.", "url": "https://wpnews.pro/news/lapo-leave-one-turn-attribution-for-self-generated-process-rewards-in-multi-turn", "canonical_source": "https://arxiv.org/abs/2607.13501", "published_at": "2026-07-16 04:00:00+00:00", "updated_at": "2026-07-16 04:31:31.018439+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "natural-language-processing", "ai-research", "ai-agents"], "entities": ["LAPO", "IGPO"], "alternates": {"html": "https://wpnews.pro/news/lapo-leave-one-turn-attribution-for-self-generated-process-rewards-in-multi-turn", "markdown": "https://wpnews.pro/news/lapo-leave-one-turn-attribution-for-self-generated-process-rewards-in-multi-turn.md", "text": "https://wpnews.pro/news/lapo-leave-one-turn-attribution-for-self-generated-process-rewards-in-multi-turn.txt", "jsonld": "https://wpnews.pro/news/lapo-leave-one-turn-attribution-for-self-generated-process-rewards-in-multi-turn.jsonld"}}