{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:OLBY5FDUE5HIQNCKA3E2AWZIIR","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":"b439973371a7088a8c826d686f23ee531eed1627abe2f31744fa6508a976f03b","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-28T14:07:48Z","title_canon_sha256":"fec4a20ac60bf55293b6e3d5812e45d9bb1ad1127781d285889ad7f64202510a"},"schema_version":"1.0","source":{"id":"2502.21074","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.21074","created_at":"2026-05-17T23:38:12Z"},{"alias_kind":"arxiv_version","alias_value":"2502.21074v3","created_at":"2026-05-17T23:38:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.21074","created_at":"2026-05-17T23:38:12Z"},{"alias_kind":"pith_short_12","alias_value":"OLBY5FDUE5HI","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"OLBY5FDUE5HIQNCK","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"OLBY5FDU","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:61fc8e220abaff7b6d2bb683a667b51d0b338e42f0a0abd112befa3e079743a6","target":"graph","created_at":"2026-05-17T23:38:12Z","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":"CODI is the first implicit CoT approach to match the performance of explicit CoT on GSM8k at the GPT-2 scale, achieving a 3.1x compression rate and outperforming the previous state-of-the-art by 28.2% in accuracy."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That aligning the hidden states of a single designated token between the explicit teacher and implicit student is sufficient to transfer full reasoning capability without loss or distortion."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"CODI compresses explicit CoT into continuous space via self-distillation and is the first implicit method to match explicit CoT performance on GSM8k at GPT-2 scale with 3.1x compression and 28.2% higher accuracy than prior implicit approaches."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Self-distillation aligns one token's hidden state to transfer chain-of-thought reasoning into continuous space without accuracy loss."}],"snapshot_sha256":"c151f3739eb36f6518d67f76896ef165a11c14218227899eb7a34ea3c42e936b"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Chain-of-Thought (CoT) reasoning enhances Large Language Models (LLMs) by encouraging step-by-step reasoning in natural language. However, leveraging a latent continuous space for reasoning may offer benefits in terms of both efficiency and robustness. Prior implicit CoT methods attempt to bypass language completely by reasoning in continuous space but have consistently underperformed compared to the standard explicit CoT approach. We introduce CODI (Continuous Chain-of-Thought via Self-Distillation), a novel training framework that effectively compresses natural language CoT into continuous s","authors_text":"Hanqi Yan, Linhai Zhang, Yali Du, Yulan He, Zhanghao Hu, Zhenyi Shen","cross_cats":[],"headline":"Self-distillation aligns one token's hidden state to transfer chain-of-thought reasoning into continuous space without accuracy loss.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-28T14:07:48Z","title":"CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation"},"references":{"count":118,"internal_anchors":7,"resolved_work":118,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Training language models to follow instructions with human feedback , url =","work_id":"04f44026-8d59-44c7-9381-031151587087","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Aho and Jeffrey D","work_id":"b1f5cb43-a3c7-4ea0-85e7-9ccc9dfe1588","year":1972},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Publications Manual , year = \"1983\", publisher =","work_id":"aca2b566-99e0-4ebb-9c7a-a81219531259","year":1983},{"cited_arxiv_id":"","doi":"10.1145/322234.322243","is_internal_anchor":false,"ref_index":4,"title":"Chandra and Dexter C","work_id":"c3270592-bd69-4213-95e1-4aaf8312be9b","year":1981},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Scalable training of","work_id":"aef70eae-f816-4598-84ec-429a2c09f5fc","year":null}],"snapshot_sha256":"61509a0e5398a54ba1b20dd6a610a19ad1c485f388ec00878d332ff503f9f93e"},"source":{"id":"2502.21074","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-17T23:11:08.279207Z","id":"2a28c913-7259-4a74-ac35-712fd648cffd","model_set":{"reader":"grok-4.3"},"one_line_summary":"CODI compresses explicit CoT into continuous space via self-distillation and is the first implicit method to match explicit CoT performance on GSM8k at GPT-2 scale with 3.1x compression and 28.2% higher accuracy than prior implicit approaches.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Self-distillation aligns one token's hidden state to transfer chain-of-thought reasoning into continuous space without accuracy loss.","strongest_claim":"CODI is the first implicit CoT approach to match the performance of explicit CoT on GSM8k at the GPT-2 scale, achieving a 3.1x compression rate and outperforming the previous state-of-the-art by 28.2% in accuracy.","weakest_assumption":"That aligning the hidden states of a single designated token between the explicit teacher and implicit student is sufficient to transfer full reasoning capability without loss or distortion."}},"verdict_id":"2a28c913-7259-4a74-ac35-712fd648cffd"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:04329874bc8097401a85406038fd455d488428ff559764f548d0f4e02c092185","target":"record","created_at":"2026-05-17T23:38:12Z","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":"b439973371a7088a8c826d686f23ee531eed1627abe2f31744fa6508a976f03b","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-28T14:07:48Z","title_canon_sha256":"fec4a20ac60bf55293b6e3d5812e45d9bb1ad1127781d285889ad7f64202510a"},"schema_version":"1.0","source":{"id":"2502.21074","kind":"arxiv","version":3}},"canonical_sha256":"72c38e9474274e88344a06c9a05b28446923da29442b1a19d42667b405a9396f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"72c38e9474274e88344a06c9a05b28446923da29442b1a19d42667b405a9396f","first_computed_at":"2026-05-17T23:38:12.664428Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:38:12.664428Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"oKGT2wVMLnwMThBdKAtaRxTTzgLhTAdhD/4eZ5wqanW0vBXmg/VkIHk8kqWk54cVGJmvqEJFwS/7o8Ki9qX4AA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:38:12.665240Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.21074","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:04329874bc8097401a85406038fd455d488428ff559764f548d0f4e02c092185","sha256:61fc8e220abaff7b6d2bb683a667b51d0b338e42f0a0abd112befa3e079743a6"],"state_sha256":"22576180c709e9fc8e6f1011226de96fbd480f76ee72c692bcfcf7836bb725b6"}