{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2022:VTZ5TICIEUABYUIMSOMAGOPCVB","short_pith_number":"pith:VTZ5TICI","canonical_record":{"source":{"id":"2210.03629","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-10-06T01:00:32Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"36a92e00015d3494a5a612ce5c1fbd2f7cd59398e873fa7146334eb68c5c68a0","abstract_canon_sha256":"cb2efab70cf9daa080135aadd32ae2c1c2aecb7f7b922c67245ba7078930e276"},"schema_version":"1.0"},"canonical_sha256":"acf3d9a04825001c510c93980339e2a851b4a95f44d0d058edccb14c85e05643","source":{"kind":"arxiv","id":"2210.03629","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2210.03629","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"arxiv_version","alias_value":"2210.03629v3","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.03629","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"pith_short_12","alias_value":"VTZ5TICIEUAB","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"pith_short_16","alias_value":"VTZ5TICIEUABYUIM","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"pith_short_8","alias_value":"VTZ5TICI","created_at":"2026-07-05T05:49:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2022:VTZ5TICIEUABYUIMSOMAGOPCVB","target":"record","payload":{"canonical_record":{"source":{"id":"2210.03629","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-10-06T01:00:32Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"36a92e00015d3494a5a612ce5c1fbd2f7cd59398e873fa7146334eb68c5c68a0","abstract_canon_sha256":"cb2efab70cf9daa080135aadd32ae2c1c2aecb7f7b922c67245ba7078930e276"},"schema_version":"1.0"},"canonical_sha256":"acf3d9a04825001c510c93980339e2a851b4a95f44d0d058edccb14c85e05643","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:49:43.463738Z","signature_b64":"zEz6r7gSnUZ8SGjw+HgeSLaFi5hUcxdW7cp0ibfGQRh4haNJkmm7Wheaslc01hPep999+JxPVuXD2fFRsRPaAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"acf3d9a04825001c510c93980339e2a851b4a95f44d0d058edccb14c85e05643","last_reissued_at":"2026-07-05T05:49:43.463381Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:49:43.463381Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2210.03629","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:49:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lP5YlP2WtqsmbRxpoJRIw1RVjk6b762poC159FVP+OWQgT9RrvLdsGKJJpa4LtCDmkKr81GiLwC2AtfbiYNIBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-17T02:05:45.646342Z"},"content_sha256":"f6a55aa66128b0ca06e6b31ddebdf55c28b114686245c7431388f55408763fc9","schema_version":"1.0","event_id":"sha256:f6a55aa66128b0ca06e6b31ddebdf55c28b114686245c7431388f55408763fc9"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2022:VTZ5TICIEUABYUIMSOMAGOPCVB","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"ReAct: Synergizing Reasoning and Acting in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A language model that writes its reasoning into the same stream as its actions plans, retrieves, and recovers from mistakes better than one that does either alone.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Dian Yu, Izhak Shafran, Jeffrey Zhao, Karthik Narasimhan, Nan Du, Shunyu Yao, Yuan Cao","submitted_at":"2022-10-06T01:00:32Z","abstract_excerpt":"While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Prompting a frozen LLM (PaLM-540B, also GPT-3) to emit interleaved free-form reasoning traces and domain actions in one trajectory outperforms reasoning-only (CoT/CoT-SC) and acting-only baselines, and on ALFWorld and WebShop with only 1–2 in-context demonstrations beats imitation and imitation+RL agents trained on 10³–10⁵ task instances by 34 and 10 absolute success-rate points, respectively.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the gains are attributable to the ReAct prompting pattern rather than to prompt-engineering and example-selection effects on a small set of held-out tasks. Prompts are hand-authored per task type, ALFWorld results are reported as best-of-6 prompt permutations against a beam-search BUTLER baseline, and HotpotQA EM (27.4) is below CoT (29.4) and far below supervised SoTA, so the headline \"outperforms\" claim depends on which benchmark and which baseline. The decision-task comparison also conflates a 540B LLM with much smaller imitation/RL agents, leaving the contribution of scale vs. method partly unidentified.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Interleaving generated reasoning traces with tool/environment actions in a single LLM prompt yields large gains over chain-of-thought-only and action-only baselines on QA, fact-checking, and text-game/web-shopping benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A language model that writes its reasoning into the same stream as its actions plans, retrieves, and recovers from mistakes better than one that does either alone.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"63324c7796d7b967b7dbd87f1d1996f9c1713fbc2813d9bf2b38e945ab01f3ea"},"source":{"id":"2210.03629","kind":"arxiv","version":3},"verdict":{"id":"190536a1-cc0c-4911-a318-e609e575d677","model_set":{"reader":"claude-opus-4-7"},"created_at":"2026-05-09T01:01:59.069339Z","strongest_claim":"Prompting a frozen LLM (PaLM-540B, also GPT-3) to emit interleaved free-form reasoning traces and domain actions in one trajectory outperforms reasoning-only (CoT/CoT-SC) and acting-only baselines, and on ALFWorld and WebShop with only 1–2 in-context demonstrations beats imitation and imitation+RL agents trained on 10³–10⁵ task instances by 34 and 10 absolute success-rate points, respectively.","one_line_summary":"Interleaving generated reasoning traces with tool/environment actions in a single LLM prompt yields large gains over chain-of-thought-only and action-only baselines on QA, fact-checking, and text-game/web-shopping benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the gains are attributable to the ReAct prompting pattern rather than to prompt-engineering and example-selection effects on a small set of held-out tasks. Prompts are hand-authored per task type, ALFWorld results are reported as best-of-6 prompt permutations against a beam-search BUTLER baseline, and HotpotQA EM (27.4) is below CoT (29.4) and far below supervised SoTA, so the headline \"outperforms\" claim depends on which benchmark and which baseline. The decision-task comparison also conflates a 540B LLM with much smaller imitation/RL agents, leaving the contribution of scale vs. method partly unidentified.","pith_extraction_headline":"A language model that writes its reasoning into the same stream as its actions plans, retrieves, and recovers from mistakes better than one that does either alone."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2210.03629/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":2,"snapshot_sha256":"93dae271c18c8146d80d7074a4cb069c4702802bb82ff048000e6627778c2920"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"190536a1-cc0c-4911-a318-e609e575d677"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T05:49:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+dM8EAKuwlRNCu2Fz9QM2bRgD+XI3nX11SRoIGQIHgGwQHB5grjl1na+mk/NKFZKHf43oaulxICQMfQowPLnAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-17T02:05:45.646811Z"},"content_sha256":"963b614fdc9e4da739ac69c1cfa4e9e59ca097e6325830145fca028394d5d1ff","schema_version":"1.0","event_id":"sha256:963b614fdc9e4da739ac69c1cfa4e9e59ca097e6325830145fca028394d5d1ff"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/VTZ5TICIEUABYUIMSOMAGOPCVB/bundle.json","state_url":"https://pith.science/pith/VTZ5TICIEUABYUIMSOMAGOPCVB/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/VTZ5TICIEUABYUIMSOMAGOPCVB/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-17T02:05:45Z","links":{"resolver":"https://pith.science/pith/VTZ5TICIEUABYUIMSOMAGOPCVB","bundle":"https://pith.science/pith/VTZ5TICIEUABYUIMSOMAGOPCVB/bundle.json","state":"https://pith.science/pith/VTZ5TICIEUABYUIMSOMAGOPCVB/state.json","well_known_bundle":"https://pith.science/.well-known/pith/VTZ5TICIEUABYUIMSOMAGOPCVB/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2022:VTZ5TICIEUABYUIMSOMAGOPCVB","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"cb2efab70cf9daa080135aadd32ae2c1c2aecb7f7b922c67245ba7078930e276","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-10-06T01:00:32Z","title_canon_sha256":"36a92e00015d3494a5a612ce5c1fbd2f7cd59398e873fa7146334eb68c5c68a0"},"schema_version":"1.0","source":{"id":"2210.03629","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2210.03629","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"arxiv_version","alias_value":"2210.03629v3","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.03629","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"pith_short_12","alias_value":"VTZ5TICIEUAB","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"pith_short_16","alias_value":"VTZ5TICIEUABYUIM","created_at":"2026-07-05T05:49:43Z"},{"alias_kind":"pith_short_8","alias_value":"VTZ5TICI","created_at":"2026-07-05T05:49:43Z"}],"graph_snapshots":[{"event_id":"sha256:963b614fdc9e4da739ac69c1cfa4e9e59ca097e6325830145fca028394d5d1ff","target":"graph","created_at":"2026-07-05T05:49:43Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Prompting a frozen LLM (PaLM-540B, also GPT-3) to emit interleaved free-form reasoning traces and domain actions in one trajectory outperforms reasoning-only (CoT/CoT-SC) and acting-only baselines, and on ALFWorld and WebShop with only 1–2 in-context demonstrations beats imitation and imitation+RL agents trained on 10³–10⁵ task instances by 34 and 10 absolute success-rate points, respectively."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the gains are attributable to the ReAct prompting pattern rather than to prompt-engineering and example-selection effects on a small set of held-out tasks. Prompts are hand-authored per task type, ALFWorld results are reported as best-of-6 prompt permutations against a beam-search BUTLER baseline, and HotpotQA EM (27.4) is below CoT (29.4) and far below supervised SoTA, so the headline \"outperforms\" claim depends on which benchmark and which baseline. The decision-task comparison also conflates a 540B LLM with much smaller imitation/RL agents, leaving the contribution of scale vs. method partly unidentified."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Interleaving generated reasoning traces with tool/environment actions in a single LLM prompt yields large gains over chain-of-thought-only and action-only baselines on QA, fact-checking, and text-game/web-shopping benchmarks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A language model that writes its reasoning into the same stream as its actions plans, retrieves, and recovers from mistakes better than one that does either alone."}],"snapshot_sha256":"63324c7796d7b967b7dbd87f1d1996f9c1713fbc2813d9bf2b38e945ab01f3ea"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"93dae271c18c8146d80d7074a4cb069c4702802bb82ff048000e6627778c2920"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2210.03629/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"While large language models (LLMs) have demonstrated impressive capabilities across tasks in language understanding and interactive decision making, their abilities for reasoning (e.g. chain-of-thought prompting) and acting (e.g. action plan generation) have primarily been studied as separate topics. In this paper, we explore the use of LLMs to generate both reasoning traces and task-specific actions in an interleaved manner, allowing for greater synergy between the two: reasoning traces help the model induce, track, and update action plans as well as handle exceptions, while actions allow it ","authors_text":"Dian Yu, Izhak Shafran, Jeffrey Zhao, Karthik Narasimhan, Nan Du, Shunyu Yao, Yuan Cao","cross_cats":["cs.AI","cs.LG"],"headline":"A language model that writes its reasoning into the same stream as its actions plans, retrieves, and recovers from mistakes better than one that does either alone.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-10-06T01:00:32Z","title":"ReAct: Synergizing Reasoning and Acting in Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.03629","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-09T01:01:59.069339Z","id":"190536a1-cc0c-4911-a318-e609e575d677","model_set":{"reader":"claude-opus-4-7"},"one_line_summary":"Interleaving generated reasoning traces with tool/environment actions in a single LLM prompt yields large gains over chain-of-thought-only and action-only baselines on QA, fact-checking, and text-game/web-shopping benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A language model that writes its reasoning into the same stream as its actions plans, retrieves, and recovers from mistakes better than one that does either alone.","strongest_claim":"Prompting a frozen LLM (PaLM-540B, also GPT-3) to emit interleaved free-form reasoning traces and domain actions in one trajectory outperforms reasoning-only (CoT/CoT-SC) and acting-only baselines, and on ALFWorld and WebShop with only 1–2 in-context demonstrations beats imitation and imitation+RL agents trained on 10³–10⁵ task instances by 34 and 10 absolute success-rate points, respectively.","weakest_assumption":"That the gains are attributable to the ReAct prompting pattern rather than to prompt-engineering and example-selection effects on a small set of held-out tasks. Prompts are hand-authored per task type, ALFWorld results are reported as best-of-6 prompt permutations against a beam-search BUTLER baseline, and HotpotQA EM (27.4) is below CoT (29.4) and far below supervised SoTA, so the headline \"outperforms\" claim depends on which benchmark and which baseline. The decision-task comparison also conflates a 540B LLM with much smaller imitation/RL agents, leaving the contribution of scale vs. method partly unidentified."}},"verdict_id":"190536a1-cc0c-4911-a318-e609e575d677"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f6a55aa66128b0ca06e6b31ddebdf55c28b114686245c7431388f55408763fc9","target":"record","created_at":"2026-07-05T05:49:43Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"cb2efab70cf9daa080135aadd32ae2c1c2aecb7f7b922c67245ba7078930e276","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-10-06T01:00:32Z","title_canon_sha256":"36a92e00015d3494a5a612ce5c1fbd2f7cd59398e873fa7146334eb68c5c68a0"},"schema_version":"1.0","source":{"id":"2210.03629","kind":"arxiv","version":3}},"canonical_sha256":"acf3d9a04825001c510c93980339e2a851b4a95f44d0d058edccb14c85e05643","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"acf3d9a04825001c510c93980339e2a851b4a95f44d0d058edccb14c85e05643","first_computed_at":"2026-07-05T05:49:43.463381Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T05:49:43.463381Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"zEz6r7gSnUZ8SGjw+HgeSLaFi5hUcxdW7cp0ibfGQRh4haNJkmm7Wheaslc01hPep999+JxPVuXD2fFRsRPaAw==","signature_status":"signed_v1","signed_at":"2026-07-05T05:49:43.463738Z","signed_message":"canonical_sha256_bytes"},"source_id":"2210.03629","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f6a55aa66128b0ca06e6b31ddebdf55c28b114686245c7431388f55408763fc9","sha256:963b614fdc9e4da739ac69c1cfa4e9e59ca097e6325830145fca028394d5d1ff"],"state_sha256":"1731f7938c8b479e4d70b4d62ec7d4a2c5402cfe85426229a7b3f901d563f4e0"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AvAfbTGduPCXe6YZrQQqF4+stpoEkxvXL3N2sjAvaTpPiOHZ8Jp6n9GS7RdOKAt7ZpjPGXDbCH5IcpJ27YdwDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-17T02:05:45.649242Z","bundle_sha256":"cd7b233ec7c31eda7b645c1ae44034b09ed135f0975774d9a5bf587c90bd9c53"}}