{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GZ3TWWSP2WZN25KOFKNRTEVFY2","short_pith_number":"pith:GZ3TWWSP","schema_version":"1.0","canonical_sha256":"36773b5a4fd5b2dd754e2a9b1992a5c6bbc91683b1809fbe4fa6e3f21c8159fe","source":{"kind":"arxiv","id":"2605.15113","version":1},"attestation_state":"computed","paper":{"title":"Learning from Language Feedback via Variational Policy Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Erik Nijkamp, Semih Yavuz, Shafiq Rayhan Joty, Yang Li","submitted_at":"2026-05-14T17:27:34Z","abstract_excerpt":"Reinforcement learning from verifiable rewards (RLVR) suffers from sparse outcome signals, creating severe exploration bottlenecks on complex reasoning tasks. Recent on-policy self-distillation methods attempt to address this by utilizing language feedback to generate dense, token-level supervision. However, these approaches rely on a fixed, passive teacher to interpret the feedback. As the student policy improves, the teacher's zero-shot assessment capabilities plateau, ultimately halting further learning. To overcome this, we propose Variational Policy Distillation (VPD), a framework that fo"},"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":"2605.15113","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T17:27:34Z","cross_cats_sorted":[],"title_canon_sha256":"5ed6f1368d548a40ee9b7d1dd9162232366a7d90110c0699e79f3d783101a5ad","abstract_canon_sha256":"3ee0e6364595354785f891512d8ebb377164dbfd468f45e7e764fcd0a44ff555"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:54.092758Z","signature_b64":"yiZ5MVtStHzFBG9tJwKgBv9B66yN7XtAF7XlLgoCKFHQQD+epnathLf50wBz1DMqETEV/o5t9m9L3hgojqvdAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"36773b5a4fd5b2dd754e2a9b1992a5c6bbc91683b1809fbe4fa6e3f21c8159fe","last_reissued_at":"2026-05-17T23:38:54.092205Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:54.092205Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning from Language Feedback via Variational Policy Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Erik Nijkamp, Semih Yavuz, Shafiq Rayhan Joty, Yang Li","submitted_at":"2026-05-14T17:27:34Z","abstract_excerpt":"Reinforcement learning from verifiable rewards (RLVR) suffers from sparse outcome signals, creating severe exploration bottlenecks on complex reasoning tasks. Recent on-policy self-distillation methods attempt to address this by utilizing language feedback to generate dense, token-level supervision. However, these approaches rely on a fixed, passive teacher to interpret the feedback. As the student policy improves, the teacher's zero-shot assessment capabilities plateau, ultimately halting further learning. To overcome this, we propose Variational Policy Distillation (VPD), a framework that fo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.15113","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"2605.15113","created_at":"2026-05-17T23:38:54.092332+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15113v1","created_at":"2026-05-17T23:38:54.092332+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15113","created_at":"2026-05-17T23:38:54.092332+00:00"},{"alias_kind":"pith_short_12","alias_value":"GZ3TWWSP2WZN","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"GZ3TWWSP2WZN25KO","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"GZ3TWWSP","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2","json":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2.json","graph_json":"https://pith.science/api/pith-number/GZ3TWWSP2WZN25KOFKNRTEVFY2/graph.json","events_json":"https://pith.science/api/pith-number/GZ3TWWSP2WZN25KOFKNRTEVFY2/events.json","paper":"https://pith.science/paper/GZ3TWWSP"},"agent_actions":{"view_html":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2","download_json":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2.json","view_paper":"https://pith.science/paper/GZ3TWWSP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15113&json=true","fetch_graph":"https://pith.science/api/pith-number/GZ3TWWSP2WZN25KOFKNRTEVFY2/graph.json","fetch_events":"https://pith.science/api/pith-number/GZ3TWWSP2WZN25KOFKNRTEVFY2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2/action/storage_attestation","attest_author":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2/action/author_attestation","sign_citation":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2/action/citation_signature","submit_replication":"https://pith.science/pith/GZ3TWWSP2WZN25KOFKNRTEVFY2/action/replication_record"}},"created_at":"2026-05-17T23:38:54.092332+00:00","updated_at":"2026-05-17T23:38:54.092332+00:00"}