{"paper":{"title":"Strat-Reasoner: Reinforcing Strategic Reasoning of LLMs in Multi-Agent Games","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Strat-Reasoner improves LLM performance in multi-agent games by 22.1 percent through recursive modeling of other agents' reasoning and group-relative reinforcement learning.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Jiarui Gan, Jiexin Wang, Mengchen Zhao, Pengxu Yang, Yi Cai, Yidong He, Yutao Lai","submitted_at":"2026-05-06T13:35:14Z","abstract_excerpt":"While Large Language Models (LLMs) excel in certain reasoning tasks, they struggle in multi-agent games where the final outcome depends on the joint strategies of all agents. In multi-agent games, the non-stationarity of other agents brings significant challenges on the evaluation of the reasoning process and the credit assignment over multiple reasoning steps. Existing single-agent reinforcement learning (RL) approaches and their multi-agent extensions fail to address these challenges as they do not incorporate other agents in the reasoning process. In this work, we propose Strat-Reasoner, a "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experimental results show that Strat-Reasoner substantially improves strategic abilities of underlying LLMs, achieving 22.1% average performance improvements across various multi-agent games.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the centralized Chain-of-Thought comparison module supplies unbiased, effective reward signals for intermediate reasoning sequences without introducing evaluation circularity or requiring hand-tuned thresholds that favor the proposed method.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Strat-Reasoner boosts LLMs' strategic performance in multi-agent games by 22.1% on average via recursive reasoning that includes other agents' thought processes and a centralized CoT evaluator for intermediate rewards.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Strat-Reasoner improves LLM performance in multi-agent games by 22.1 percent through recursive modeling of other agents' reasoning and group-relative reinforcement learning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b5ae241aac07cfe364bced591e61ca5dc97c507790b8fad1d8146427c0b5b4a8"},"source":{"id":"2605.04906","kind":"arxiv","version":2},"verdict":{"id":"85f771d2-7a4b-49bc-8922-e36bcc331a7d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T18:25:47.205301Z","strongest_claim":"Experimental results show that Strat-Reasoner substantially improves strategic abilities of underlying LLMs, achieving 22.1% average performance improvements across various multi-agent games.","one_line_summary":"Strat-Reasoner boosts LLMs' strategic performance in multi-agent games by 22.1% on average via recursive reasoning that includes other agents' thought processes and a centralized CoT evaluator for intermediate rewards.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the centralized Chain-of-Thought comparison module supplies unbiased, effective reward signals for intermediate reasoning sequences without introducing evaluation circularity or requiring hand-tuned thresholds that favor the proposed method.","pith_extraction_headline":"Strat-Reasoner improves LLM performance in multi-agent games by 22.1 percent through recursive modeling of other agents' reasoning and group-relative reinforcement learning."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.04906/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T10:40:58.678339Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T22:01:23.639665Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T14:03:34.797473Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"660aa19132ef432a81225a326119d4dff6b9e4c7ef9e51077070810613b887bc"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"12137ea112bcb4f77d1a5600f9be83d309dfbf5d8d925625990081c1ecdbd74a"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}