{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:R7WGQ7MHUN6HU3TDDO6NJLH66G","short_pith_number":"pith:R7WGQ7MH","schema_version":"1.0","canonical_sha256":"8fec687d87a37c7a6e631bbcd4acfef1adaf747a701ef48f0332fc3416df1d73","source":{"kind":"arxiv","id":"2407.06023","version":3},"attestation_state":"computed","paper":{"title":"Distilling System 2 into System 1","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ilia Kulikov, Jason Weston, Jing Xu, Ping Yu","submitted_at":"2024-07-08T15:17:46Z","abstract_excerpt":"Large language models (LLMs) can spend extra compute during inference to generate intermediate thoughts, which helps to produce better final responses. Since Chain-of-Thought (Wei et al., 2022), many such System 2 techniques have been proposed such as Rephrase and Respond (Deng et al., 2023a), System 2 Attention (Weston and Sukhbaatar, 2023) and Branch-Solve-Merge (Saha et al., 2023). In this work we investigate self-supervised methods to ``compile'' (distill) higher quality outputs from System 2 techniques back into LLM generations without intermediate reasoning token sequences, as this reaso"},"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":"2407.06023","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-08T15:17:46Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2c85c7b39d8054d45140cabf75da92cd5e1e40ce00d3e56e112488f43f117679","abstract_canon_sha256":"3ec35f44c682578b490638931238c6b1b5dccb9c49bf0a2c7806d807322d5026"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:48:15.400979Z","signature_b64":"2Ri1BOHywS7BqJxEYlw8V1OD6sb04EznXYxVUglSjoqb+Q+xHzPbYFeDwWRdHWpBzuTuDL13Qj83AMIWl49CCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8fec687d87a37c7a6e631bbcd4acfef1adaf747a701ef48f0332fc3416df1d73","last_reissued_at":"2026-07-05T08:48:15.400388Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:48:15.400388Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Distilling System 2 into System 1","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ilia Kulikov, Jason Weston, Jing Xu, Ping Yu","submitted_at":"2024-07-08T15:17:46Z","abstract_excerpt":"Large language models (LLMs) can spend extra compute during inference to generate intermediate thoughts, which helps to produce better final responses. Since Chain-of-Thought (Wei et al., 2022), many such System 2 techniques have been proposed such as Rephrase and Respond (Deng et al., 2023a), System 2 Attention (Weston and Sukhbaatar, 2023) and Branch-Solve-Merge (Saha et al., 2023). In this work we investigate self-supervised methods to ``compile'' (distill) higher quality outputs from System 2 techniques back into LLM generations without intermediate reasoning token sequences, as this reaso"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.06023","kind":"arxiv","version":3},"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/2407.06023/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":"2407.06023","created_at":"2026-07-05T08:48:15.400491+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.06023v3","created_at":"2026-07-05T08:48:15.400491+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.06023","created_at":"2026-07-05T08:48:15.400491+00:00"},{"alias_kind":"pith_short_12","alias_value":"R7WGQ7MHUN6H","created_at":"2026-07-05T08:48:15.400491+00:00"},{"alias_kind":"pith_short_16","alias_value":"R7WGQ7MHUN6HU3TD","created_at":"2026-07-05T08:48:15.400491+00:00"},{"alias_kind":"pith_short_8","alias_value":"R7WGQ7MH","created_at":"2026-07-05T08:48:15.400491+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27268","citing_title":"E-TTS: A New Embodied Test-Time Scaling Framework for Robotic Manipulation","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2607.02234","citing_title":"Purified OPSD: On-Policy Self-Distillation Without Losing How to Think","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03503","citing_title":"ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17890","citing_title":"Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2509.05489","citing_title":"Self-Aligned Reward: Towards Effective and Efficient Reasoners","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2601.11340","citing_title":"Neural Chain-of-Thought Search: Searching the Optimal Reasoning Path to Enhance Large Language Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2503.16419","citing_title":"Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models","ref_index":219,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08221","citing_title":"NoisyCoconut: Counterfactual Consensus via Latent Space Reasoning","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25110","citing_title":"Knowledge Distillation Must Account for What It Loses","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25110","citing_title":"Knowledge Distillation Must Account for What It Loses","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06165","citing_title":"Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10674","citing_title":"Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2412.06769","citing_title":"Training Large Language Models to Reason in a Continuous Latent Space","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17304","citing_title":"Efficient Test-Time Scaling via Temporal Reasoning Aggregation","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G","json":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G.json","graph_json":"https://pith.science/api/pith-number/R7WGQ7MHUN6HU3TDDO6NJLH66G/graph.json","events_json":"https://pith.science/api/pith-number/R7WGQ7MHUN6HU3TDDO6NJLH66G/events.json","paper":"https://pith.science/paper/R7WGQ7MH"},"agent_actions":{"view_html":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G","download_json":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G.json","view_paper":"https://pith.science/paper/R7WGQ7MH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.06023&json=true","fetch_graph":"https://pith.science/api/pith-number/R7WGQ7MHUN6HU3TDDO6NJLH66G/graph.json","fetch_events":"https://pith.science/api/pith-number/R7WGQ7MHUN6HU3TDDO6NJLH66G/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G/action/storage_attestation","attest_author":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G/action/author_attestation","sign_citation":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G/action/citation_signature","submit_replication":"https://pith.science/pith/R7WGQ7MHUN6HU3TDDO6NJLH66G/action/replication_record"}},"created_at":"2026-07-05T08:48:15.400491+00:00","updated_at":"2026-07-05T08:48:15.400491+00:00"}