{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:Y5BR26G7LCNH3OGEJYBPXFAPA6","short_pith_number":"pith:Y5BR26G7","canonical_record":{"source":{"id":"2605.14558","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T08:33:02Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"381fa8f479b09790d076d448166f439f82aaf7e52afc4bf0e30ededcc7af5a81","abstract_canon_sha256":"769839a6b2f431823f2ce011bc079382dda2ac288ba14a42e289655beab38918"},"schema_version":"1.0"},"canonical_sha256":"c7431d78df589a7db8c44e02fb940f07806e1939ab66a013470fbf4fc8a89aab","source":{"kind":"arxiv","id":"2605.14558","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14558","created_at":"2026-05-17T23:39:05Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14558v1","created_at":"2026-05-17T23:39:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14558","created_at":"2026-05-17T23:39:05Z"},{"alias_kind":"pith_short_12","alias_value":"Y5BR26G7LCNH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"Y5BR26G7LCNH3OGE","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"Y5BR26G7","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:Y5BR26G7LCNH3OGEJYBPXFAPA6","target":"record","payload":{"canonical_record":{"source":{"id":"2605.14558","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T08:33:02Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"381fa8f479b09790d076d448166f439f82aaf7e52afc4bf0e30ededcc7af5a81","abstract_canon_sha256":"769839a6b2f431823f2ce011bc079382dda2ac288ba14a42e289655beab38918"},"schema_version":"1.0"},"canonical_sha256":"c7431d78df589a7db8c44e02fb940f07806e1939ab66a013470fbf4fc8a89aab","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:05.628770Z","signature_b64":"AMh57IO1yLCmyvqcVWopqckg5PFtwAtn2gjTxBEjEPVoYbKxNh81jPR3AIrA8kMI7D3QyjLgiu3J5enTHosvCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c7431d78df589a7db8c44e02fb940f07806e1939ab66a013470fbf4fc8a89aab","last_reissued_at":"2026-05-17T23:39:05.628152Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:05.628152Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.14558","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:05Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vYDvUkYZIjwC0vUmZkI219O0trzn3NW4tS3ZPWZSY4BFkZ1HQVqjCUxGwYFdkhHBLGBiidwDovPbgrswrPt1Cw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T21:23:27.437717Z"},"content_sha256":"4ded5eb164d1f9c3ff776767746d4431716bd7ce53f530c6f0d5d50c25514c0d","schema_version":"1.0","event_id":"sha256:4ded5eb164d1f9c3ff776767746d4431716bd7ce53f530c6f0d5d50c25514c0d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:Y5BR26G7LCNH3OGEJYBPXFAPA6","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Resolving Action Bottleneck: Agentic Reinforcement Learning Informed by Token-Level Energy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Token-level signals concentrate on action tokens in agentic RL, so reweighting gradients toward them outperforms uniform policy gradients.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"David Wipf, Henry Peng Zou, Junhua Liu, Junyou Zhu, Langzhou He, Philip S. Yu, Qitian Wu, Wei-Chieh Huang, Yue Zhou, Zhengyao Gu","submitted_at":"2026-05-14T08:33:02Z","abstract_excerpt":"Agentic reinforcement learning trains large language models using multi-turn trajectories that interleave long reasoning traces with short environment-facing actions. Common policy-gradient methods, such as PPO and GRPO, treat each token in a trajectory equally, leading to uniform credit assignment. In this paper, we critically demonstrate that such uniform credit assignment largely misallocates token-level training signals. From an energy-based modeling perspective, we show that token-level training signals, quantified by their correlations with reward variance of different rollouts sampled f"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"token-level training signals, quantified by their correlations with reward variance of different rollouts sampled from a given prompt, concentrate sharply on action tokens rather than reasoning tokens, even though action tokens account for only a small fraction of the trajectory","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"that down-weighting reasoning tokens and boosting high-uncertainty action tokens will not degrade the quality of the reasoning chain itself or introduce new instabilities in long-horizon trajectories","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ActFocus resolves the action bottleneck in agentic RL by reweighting token gradients toward action tokens using observed reward variance and an energy-based uncertainty term, outperforming PPO and GRPO by up to 65 percentage points.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Token-level signals concentrate on action tokens in agentic RL, so reweighting gradients toward them outperforms uniform policy gradients.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"89df5c0110227a37e42f43dd209e275d1d2e766ec454baf5a00e52c833d09f07"},"source":{"id":"2605.14558","kind":"arxiv","version":1},"verdict":{"id":"0fc00fb7-6634-456c-8f63-2e929df9641c","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T01:43:12.935990Z","strongest_claim":"token-level training signals, quantified by their correlations with reward variance of different rollouts sampled from a given prompt, concentrate sharply on action tokens rather than reasoning tokens, even though action tokens account for only a small fraction of the trajectory","one_line_summary":"ActFocus resolves the action bottleneck in agentic RL by reweighting token gradients toward action tokens using observed reward variance and an energy-based uncertainty term, outperforming PPO and GRPO by up to 65 percentage points.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"that down-weighting reasoning tokens and boosting high-uncertainty action tokens will not degrade the quality of the reasoning chain itself or introduce new instabilities in long-horizon trajectories","pith_extraction_headline":"Token-level signals concentrate on action tokens in agentic RL, so reweighting gradients toward them outperforms uniform policy gradients."},"references":{"count":39,"sample":[{"doi":"","year":2023,"title":"Lmrl gym: Benchmarks for multi-turn reinforcement learn- ing with language models","work_id":"bad1baad-5c3f-4456-ac96-29c8f5e78bfb","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Chan, Hao Sun, Samuel Holt, and Mihaela van der Schaar","work_id":"c8748d0b-32ee-428e-a008-b6fee31a2bff","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"arXiv preprint arXiv:2502.01600 , year=","work_id":"7e929792-6a2a-42ff-a1db-763f890d8b4e","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Process reinforcement through implicit rewards,","work_id":"9ea5754d-d348-48ee-9421-97a9f4c6d3eb","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Process Reinforcement through Implicit Rewards","work_id":"c31a2126-86f9-44f3-91f3-208d0fc1463a","ref_index":5,"cited_arxiv_id":"2502.01456","is_internal_anchor":true}],"resolved_work":39,"snapshot_sha256":"8887903501c992eb3daf01e2640401d8063f10ac6db147c7fe14f36e48e00b83","internal_anchors":17},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2658e067e3a21f2e238b50c054966600406cf2f3e106331f0fd19623b5045d36"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"0fc00fb7-6634-456c-8f63-2e929df9641c"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:05Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"vOHWxRtMri6zTEXU/QBMZfkUOp4k7kRRkEd3vwsadD7m9H6TY+zt9WHrRqw9+hN8aCn4EXe+U1m1MgCzceEdBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-25T21:23:27.438760Z"},"content_sha256":"c67e1fa3a048192babe0dbf22e2afb2b87cf6223de6ba24f7b199f9d0db1934f","schema_version":"1.0","event_id":"sha256:c67e1fa3a048192babe0dbf22e2afb2b87cf6223de6ba24f7b199f9d0db1934f"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6/bundle.json","state_url":"https://pith.science/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6/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":"2026-05-25T21:23:27Z","links":{"resolver":"https://pith.science/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6","bundle":"https://pith.science/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6/bundle.json","state":"https://pith.science/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6/state.json","well_known_bundle":"https://pith.science/.well-known/pith/Y5BR26G7LCNH3OGEJYBPXFAPA6/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:Y5BR26G7LCNH3OGEJYBPXFAPA6","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":"769839a6b2f431823f2ce011bc079382dda2ac288ba14a42e289655beab38918","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T08:33:02Z","title_canon_sha256":"381fa8f479b09790d076d448166f439f82aaf7e52afc4bf0e30ededcc7af5a81"},"schema_version":"1.0","source":{"id":"2605.14558","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14558","created_at":"2026-05-17T23:39:05Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14558v1","created_at":"2026-05-17T23:39:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14558","created_at":"2026-05-17T23:39:05Z"},{"alias_kind":"pith_short_12","alias_value":"Y5BR26G7LCNH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"Y5BR26G7LCNH3OGE","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"Y5BR26G7","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:c67e1fa3a048192babe0dbf22e2afb2b87cf6223de6ba24f7b199f9d0db1934f","target":"graph","created_at":"2026-05-17T23:39:05Z","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":"token-level training signals, quantified by their correlations with reward variance of different rollouts sampled from a given prompt, concentrate sharply on action tokens rather than reasoning tokens, even though action tokens account for only a small fraction of the trajectory"},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"that down-weighting reasoning tokens and boosting high-uncertainty action tokens will not degrade the quality of the reasoning chain itself or introduce new instabilities in long-horizon trajectories"},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ActFocus resolves the action bottleneck in agentic RL by reweighting token gradients toward action tokens using observed reward variance and an energy-based uncertainty term, outperforming PPO and GRPO by up to 65 percentage points."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Token-level signals concentrate on action tokens in agentic RL, so reweighting gradients toward them outperforms uniform policy gradients."}],"snapshot_sha256":"89df5c0110227a37e42f43dd209e275d1d2e766ec454baf5a00e52c833d09f07"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2658e067e3a21f2e238b50c054966600406cf2f3e106331f0fd19623b5045d36"},"paper":{"abstract_excerpt":"Agentic reinforcement learning trains large language models using multi-turn trajectories that interleave long reasoning traces with short environment-facing actions. Common policy-gradient methods, such as PPO and GRPO, treat each token in a trajectory equally, leading to uniform credit assignment. In this paper, we critically demonstrate that such uniform credit assignment largely misallocates token-level training signals. From an energy-based modeling perspective, we show that token-level training signals, quantified by their correlations with reward variance of different rollouts sampled f","authors_text":"David Wipf, Henry Peng Zou, Junhua Liu, Junyou Zhu, Langzhou He, Philip S. Yu, Qitian Wu, Wei-Chieh Huang, Yue Zhou, Zhengyao Gu","cross_cats":["cs.AI","cs.CL"],"headline":"Token-level signals concentrate on action tokens in agentic RL, so reweighting gradients toward them outperforms uniform policy gradients.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T08:33:02Z","title":"Resolving Action Bottleneck: Agentic Reinforcement Learning Informed by Token-Level Energy"},"references":{"count":39,"internal_anchors":17,"resolved_work":39,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Lmrl gym: Benchmarks for multi-turn reinforcement learn- ing with language models","work_id":"bad1baad-5c3f-4456-ac96-29c8f5e78bfb","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Chan, Hao Sun, Samuel Holt, and Mihaela van der Schaar","work_id":"c8748d0b-32ee-428e-a008-b6fee31a2bff","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"arXiv preprint arXiv:2502.01600 , year=","work_id":"7e929792-6a2a-42ff-a1db-763f890d8b4e","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Process reinforcement through implicit rewards,","work_id":"9ea5754d-d348-48ee-9421-97a9f4c6d3eb","year":null},{"cited_arxiv_id":"2502.01456","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Process Reinforcement through Implicit Rewards","work_id":"c31a2126-86f9-44f3-91f3-208d0fc1463a","year":null}],"snapshot_sha256":"8887903501c992eb3daf01e2640401d8063f10ac6db147c7fe14f36e48e00b83"},"source":{"id":"2605.14558","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T01:43:12.935990Z","id":"0fc00fb7-6634-456c-8f63-2e929df9641c","model_set":{"reader":"grok-4.3"},"one_line_summary":"ActFocus resolves the action bottleneck in agentic RL by reweighting token gradients toward action tokens using observed reward variance and an energy-based uncertainty term, outperforming PPO and GRPO by up to 65 percentage points.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Token-level signals concentrate on action tokens in agentic RL, so reweighting gradients toward them outperforms uniform policy gradients.","strongest_claim":"token-level training signals, quantified by their correlations with reward variance of different rollouts sampled from a given prompt, concentrate sharply on action tokens rather than reasoning tokens, even though action tokens account for only a small fraction of the trajectory","weakest_assumption":"that down-weighting reasoning tokens and boosting high-uncertainty action tokens will not degrade the quality of the reasoning chain itself or introduce new instabilities in long-horizon trajectories"}},"verdict_id":"0fc00fb7-6634-456c-8f63-2e929df9641c"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:4ded5eb164d1f9c3ff776767746d4431716bd7ce53f530c6f0d5d50c25514c0d","target":"record","created_at":"2026-05-17T23:39:05Z","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":"769839a6b2f431823f2ce011bc079382dda2ac288ba14a42e289655beab38918","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T08:33:02Z","title_canon_sha256":"381fa8f479b09790d076d448166f439f82aaf7e52afc4bf0e30ededcc7af5a81"},"schema_version":"1.0","source":{"id":"2605.14558","kind":"arxiv","version":1}},"canonical_sha256":"c7431d78df589a7db8c44e02fb940f07806e1939ab66a013470fbf4fc8a89aab","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c7431d78df589a7db8c44e02fb940f07806e1939ab66a013470fbf4fc8a89aab","first_computed_at":"2026-05-17T23:39:05.628152Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:05.628152Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"AMh57IO1yLCmyvqcVWopqckg5PFtwAtn2gjTxBEjEPVoYbKxNh81jPR3AIrA8kMI7D3QyjLgiu3J5enTHosvCw==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:05.628770Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14558","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:4ded5eb164d1f9c3ff776767746d4431716bd7ce53f530c6f0d5d50c25514c0d","sha256:c67e1fa3a048192babe0dbf22e2afb2b87cf6223de6ba24f7b199f9d0db1934f"],"state_sha256":"e297fa6a3570353fd98997fc35c7e37cf643a0cb1db02c28096171b41c58448f"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Ut/6fhepZH70n4DTpjPznGkWJbqEVmRWpXGWdxDGygCBCnj9Q2bsg7UqRmUl5/slwXhumRIBgiY6121kaqP5Ag==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-25T21:23:27.442988Z","bundle_sha256":"832bf4a953b767799b82a8fb59d9f4dd3b4766030f78f4cf811b99ce8810d0d2"}}