{"paper":{"title":"Bridging MARL to SARL: An Order-Independent Multi-Agent Transformer via Latent Consensus","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"CMAT reformulates cooperative MARL as hierarchical SARL by generating a latent consensus vector that conditions simultaneous order-independent actions.","cross_cats":["cs.AI","cs.MA"],"primary_cat":"cs.LG","authors_text":"Jing Gao, Sen Li, Zijian Zhao","submitted_at":"2026-04-15T04:52:22Z","abstract_excerpt":"Cooperative multi-agent reinforcement learning (MARL) is widely used to address large joint observation and action spaces by decomposing a centralized control problem into multiple interacting agents. However, such decomposition often introduces additional challenges, including non-stationarity, unstable training, weak coordination, and limited theoretical guarantees. In this paper, we propose the Consensus Multi-Agent Transformer (CMAT), a centralized framework that bridges cooperative MARL to a hierarchical single-agent reinforcement learning (SARL) formulation. CMAT treats all agents as a u"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CMAT bridges cooperative MARL to a hierarchical SARL formulation by autoregressively generating a high-level consensus vector that conditions simultaneous, order-independent action generation, allowing optimization with single-agent PPO while achieving superior performance on StarCraft II, Multi-Agent MuJoCo, and Google Research Football benchmarks.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The latent consensus vector produced by the transformer decoder actually captures sufficient coordination information to replace explicit agent-to-agent communication or centralized critics without loss of expressivity; this is assumed rather than proven and is only supported by empirical wins on the chosen benchmarks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CMAT uses a transformer decoder to produce a high-level consensus vector in latent space, enabling simultaneous order-independent actions by all agents and optimization via single-agent PPO, with superior results on StarCraft II, Multi-Agent MuJoCo, and Google Research Football.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CMAT reformulates cooperative MARL as hierarchical SARL by generating a latent consensus vector that conditions simultaneous order-independent actions.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"32da2ca73a9024407380f65588279b58e41459d34d6de064af125d9e6a8d1187"},"source":{"id":"2604.13472","kind":"arxiv","version":2},"verdict":{"id":"8b1eb208-6b17-4a2e-815a-e7433c534e05","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T13:37:34.576192Z","strongest_claim":"CMAT bridges cooperative MARL to a hierarchical SARL formulation by autoregressively generating a high-level consensus vector that conditions simultaneous, order-independent action generation, allowing optimization with single-agent PPO while achieving superior performance on StarCraft II, Multi-Agent MuJoCo, and Google Research Football benchmarks.","one_line_summary":"CMAT uses a transformer decoder to produce a high-level consensus vector in latent space, enabling simultaneous order-independent actions by all agents and optimization via single-agent PPO, with superior results on StarCraft II, Multi-Agent MuJoCo, and Google Research Football.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The latent consensus vector produced by the transformer decoder actually captures sufficient coordination information to replace explicit agent-to-agent communication or centralized critics without loss of expressivity; this is assumed rather than proven and is only supported by empirical wins on the chosen benchmarks.","pith_extraction_headline":"CMAT reformulates cooperative MARL as hierarchical SARL by generating a latent consensus vector that conditions simultaneous order-independent actions."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.13472/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}