{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KFDYI7UA7ZVAX5F4G64AVTK6GS","short_pith_number":"pith:KFDYI7UA","schema_version":"1.0","canonical_sha256":"5147847e80fe6a0bf4bc37b80acd5e349122412773e3fb40b1acd28c4a66b52a","source":{"kind":"arxiv","id":"2605.19433","version":1},"attestation_state":"computed","paper":{"title":"Backtracking When It Strays: Mitigating Dual Exposure Biases in LLM Reasoning Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bing Wang, Chen Shen, Jieping Ye, Kaiyuan Liu, Rui Miao, Shaotian Yan, Sinan Fan, Xiaosong Yuan, Ximing Li, Zhanming Shen","submitted_at":"2026-05-19T06:42:52Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable success in complex reasoning tasks via long chain-of-thought (CoT), yet their immense computational overhead hinders real-world deployment. LLM reasoning distillation addresses this by transferring reasoning capabilities from formidable teacher models to compact student models. However, existing distillation paradigms face a fundamental dilemma. Typical off-policy distillation strictly utilizes teacher-generated golden trajectories, suffering from an exposure bias due to the mismatch between training distributions and student-generated infe"},"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":"2605.19433","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-19T06:42:52Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c9c32a45ee559d0424672e281902b8a676831d1994df7cc9962083d1bbbb313a","abstract_canon_sha256":"d4f42f67a434e4c832b74bad58836c3080335f945c3ddf14e015bdfb8310ac09"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:45.631101Z","signature_b64":"murOO/b+Efmm07qY1E4rJGDR1gEt4MM24iHm5Bt4Kmq7RCliQkzm3Y+IUAOsWUd1xPQzazOM8EBG+wR975RCDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5147847e80fe6a0bf4bc37b80acd5e349122412773e3fb40b1acd28c4a66b52a","last_reissued_at":"2026-05-20T01:05:45.630143Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:45.630143Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Backtracking When It Strays: Mitigating Dual Exposure Biases in LLM Reasoning Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bing Wang, Chen Shen, Jieping Ye, Kaiyuan Liu, Rui Miao, Shaotian Yan, Sinan Fan, Xiaosong Yuan, Ximing Li, Zhanming Shen","submitted_at":"2026-05-19T06:42:52Z","abstract_excerpt":"Large language models (LLMs) have achieved remarkable success in complex reasoning tasks via long chain-of-thought (CoT), yet their immense computational overhead hinders real-world deployment. LLM reasoning distillation addresses this by transferring reasoning capabilities from formidable teacher models to compact student models. However, existing distillation paradigms face a fundamental dilemma. Typical off-policy distillation strictly utilizes teacher-generated golden trajectories, suffering from an exposure bias due to the mismatch between training distributions and student-generated infe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19433","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.19433/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":"2605.19433","created_at":"2026-05-20T01:05:45.630309+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19433v1","created_at":"2026-05-20T01:05:45.630309+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19433","created_at":"2026-05-20T01:05:45.630309+00:00"},{"alias_kind":"pith_short_12","alias_value":"KFDYI7UA7ZVA","created_at":"2026-05-20T01:05:45.630309+00:00"},{"alias_kind":"pith_short_16","alias_value":"KFDYI7UA7ZVAX5F4","created_at":"2026-05-20T01:05:45.630309+00:00"},{"alias_kind":"pith_short_8","alias_value":"KFDYI7UA","created_at":"2026-05-20T01:05:45.630309+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/KFDYI7UA7ZVAX5F4G64AVTK6GS","json":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS.json","graph_json":"https://pith.science/api/pith-number/KFDYI7UA7ZVAX5F4G64AVTK6GS/graph.json","events_json":"https://pith.science/api/pith-number/KFDYI7UA7ZVAX5F4G64AVTK6GS/events.json","paper":"https://pith.science/paper/KFDYI7UA"},"agent_actions":{"view_html":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS","download_json":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS.json","view_paper":"https://pith.science/paper/KFDYI7UA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19433&json=true","fetch_graph":"https://pith.science/api/pith-number/KFDYI7UA7ZVAX5F4G64AVTK6GS/graph.json","fetch_events":"https://pith.science/api/pith-number/KFDYI7UA7ZVAX5F4G64AVTK6GS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS/action/storage_attestation","attest_author":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS/action/author_attestation","sign_citation":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS/action/citation_signature","submit_replication":"https://pith.science/pith/KFDYI7UA7ZVAX5F4G64AVTK6GS/action/replication_record"}},"created_at":"2026-05-20T01:05:45.630309+00:00","updated_at":"2026-05-20T01:05:45.630309+00:00"}