{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:X72JH4KJPNYWISDSWUGEOD37KV","short_pith_number":"pith:X72JH4KJ","schema_version":"1.0","canonical_sha256":"bff493f1497b71644872b50c470f7f5547e08b4d09f852470ad88d1ff1dd2094","source":{"kind":"arxiv","id":"2003.06709","version":5},"attestation_state":"computed","paper":{"title":"FACMAC: Factored Multi-Agent Centralised Policy Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bei Peng, Christian A. Schroeder de Witt, Philip H. S. Torr, Pierre-Alexandre Kamienny, Shimon Whiteson, Tabish Rashid, Wendelin B\\\"ohmer","submitted_at":"2020-03-14T21:29:09Z","abstract_excerpt":"We propose FACtored Multi-Agent Centralised policy gradients (FACMAC), a new method for cooperative multi-agent reinforcement learning in both discrete and continuous action spaces. Like MADDPG, a popular multi-agent actor-critic method, our approach uses deep deterministic policy gradients to learn policies. However, FACMAC learns a centralised but factored critic, which combines per-agent utilities into the joint action-value function via a non-linear monotonic function, as in QMIX, a popular multi-agent Q-learning algorithm. However, unlike QMIX, there are no inherent constraints on factori"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2003.06709","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-03-14T21:29:09Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"6505fbd71a2075d33e54525824d5636710e8db22dc2358b58261f7b885994128","abstract_canon_sha256":"2e91cdd2ba054061b6afbfe0d5d87f2a914127c1f46cc49db50df76925affc1f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:38:13.024713Z","signature_b64":"uYzyEcvCiDxEC1F0p3/K5nF7Lhl7l6ZIy6z0ARBbYj9EIIxo7T0/kAb2aQ6o7oW8zk/7ifl2bhLkszjKkyJuCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bff493f1497b71644872b50c470f7f5547e08b4d09f852470ad88d1ff1dd2094","last_reissued_at":"2026-07-05T02:38:13.024290Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:38:13.024290Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FACMAC: Factored Multi-Agent Centralised Policy Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bei Peng, Christian A. Schroeder de Witt, Philip H. S. Torr, Pierre-Alexandre Kamienny, Shimon Whiteson, Tabish Rashid, Wendelin B\\\"ohmer","submitted_at":"2020-03-14T21:29:09Z","abstract_excerpt":"We propose FACtored Multi-Agent Centralised policy gradients (FACMAC), a new method for cooperative multi-agent reinforcement learning in both discrete and continuous action spaces. Like MADDPG, a popular multi-agent actor-critic method, our approach uses deep deterministic policy gradients to learn policies. However, FACMAC learns a centralised but factored critic, which combines per-agent utilities into the joint action-value function via a non-linear monotonic function, as in QMIX, a popular multi-agent Q-learning algorithm. However, unlike QMIX, there are no inherent constraints on factori"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2003.06709","kind":"arxiv","version":5},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2003.06709/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2003.06709","created_at":"2026-07-05T02:38:13.024364+00:00"},{"alias_kind":"arxiv_version","alias_value":"2003.06709v5","created_at":"2026-07-05T02:38:13.024364+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2003.06709","created_at":"2026-07-05T02:38:13.024364+00:00"},{"alias_kind":"pith_short_12","alias_value":"X72JH4KJPNYW","created_at":"2026-07-05T02:38:13.024364+00:00"},{"alias_kind":"pith_short_16","alias_value":"X72JH4KJPNYWISDS","created_at":"2026-07-05T02:38:13.024364+00:00"},{"alias_kind":"pith_short_8","alias_value":"X72JH4KJ","created_at":"2026-07-05T02:38:13.024364+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25526","citing_title":"Low Variance Trust Region Optimization with Independent Actors and Sequential Updates in Cooperative Multi-agent Reinforcement Learning","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.21085","citing_title":"Sim2O: Efficient Offline-to-Online MARL via Joint Action Composition","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12655","citing_title":"Robust Instruction Compliance in Cooperative Multi-Agent Reinforcement Learning","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26286","citing_title":"Decoupled Delay Compensation: Enhancing Pre-trained MARL Policies via Learned Dynamics Filtering","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13472","citing_title":"Bridging MARL to SARL: An Order-Independent Multi-Agent Transformer via Latent Consensus","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV","json":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV.json","graph_json":"https://pith.science/api/pith-number/X72JH4KJPNYWISDSWUGEOD37KV/graph.json","events_json":"https://pith.science/api/pith-number/X72JH4KJPNYWISDSWUGEOD37KV/events.json","paper":"https://pith.science/paper/X72JH4KJ"},"agent_actions":{"view_html":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV","download_json":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV.json","view_paper":"https://pith.science/paper/X72JH4KJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2003.06709&json=true","fetch_graph":"https://pith.science/api/pith-number/X72JH4KJPNYWISDSWUGEOD37KV/graph.json","fetch_events":"https://pith.science/api/pith-number/X72JH4KJPNYWISDSWUGEOD37KV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV/action/storage_attestation","attest_author":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV/action/author_attestation","sign_citation":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV/action/citation_signature","submit_replication":"https://pith.science/pith/X72JH4KJPNYWISDSWUGEOD37KV/action/replication_record"}},"created_at":"2026-07-05T02:38:13.024364+00:00","updated_at":"2026-07-05T02:38:13.024364+00:00"}