{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KWFR4RRCVD4J6GBF3FVBWRXXXS","short_pith_number":"pith:KWFR4RRC","schema_version":"1.0","canonical_sha256":"558b1e4622a8f89f1825d96a1b46f7bc95de93d0326037cd90f6c153b6ad8220","source":{"kind":"arxiv","id":"2606.05988","version":1},"attestation_state":"computed","paper":{"title":"Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Maxime Griot, Paul Steven Scotti, Tanishq Mathew Abraham","submitted_at":"2026-06-04T10:30:58Z","abstract_excerpt":"Reasoning models produce long chain-of-thought traces that are costly to distill and encourage verbose student outputs. We study post-hoc compression of such traces before knowledge distillation. Two teachers, Qwen3.5-397B-A17B and gpt-oss-120B, generate about 283k correct traces each; two instruction-tuned models then compress them to 8.6-21.0% of their original character length. Across a 48-run main grid plus seven Qwen-teacher truncation ablations, compressed traces reduce training tokens to 12-30% of raw, speed up training by 2.0-7.6x, and shorten inference outputs by 3-19x with smaller re"},"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":"2606.05988","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-04T10:30:58Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"63c08c2a9b1a6a993f0ba4f19fcad72ff298977186a86ec70d2c92a01d720cd6","abstract_canon_sha256":"2ad944876e397c6837838baac1da278674d2497de3e3daef7b4dd4a4fad62753"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:15:29.704687Z","signature_b64":"yQgfKm6gO8k0dW5G0FTlKda6DShRlAhSmMxHmALV3VCgqhCn3FcYDiX5pQAh+Ta276W02Cou6CGY1wue3J+JBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"558b1e4622a8f89f1825d96a1b46f7bc95de93d0326037cd90f6c153b6ad8220","last_reissued_at":"2026-06-05T01:15:29.704110Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:15:29.704110Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Maxime Griot, Paul Steven Scotti, Tanishq Mathew Abraham","submitted_at":"2026-06-04T10:30:58Z","abstract_excerpt":"Reasoning models produce long chain-of-thought traces that are costly to distill and encourage verbose student outputs. We study post-hoc compression of such traces before knowledge distillation. Two teachers, Qwen3.5-397B-A17B and gpt-oss-120B, generate about 283k correct traces each; two instruction-tuned models then compress them to 8.6-21.0% of their original character length. Across a 48-run main grid plus seven Qwen-teacher truncation ablations, compressed traces reduce training tokens to 12-30% of raw, speed up training by 2.0-7.6x, and shorten inference outputs by 3-19x with smaller re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.05988","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/2606.05988/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":"2606.05988","created_at":"2026-06-05T01:15:29.704196+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.05988v1","created_at":"2026-06-05T01:15:29.704196+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.05988","created_at":"2026-06-05T01:15:29.704196+00:00"},{"alias_kind":"pith_short_12","alias_value":"KWFR4RRCVD4J","created_at":"2026-06-05T01:15:29.704196+00:00"},{"alias_kind":"pith_short_16","alias_value":"KWFR4RRCVD4J6GBF","created_at":"2026-06-05T01:15:29.704196+00:00"},{"alias_kind":"pith_short_8","alias_value":"KWFR4RRC","created_at":"2026-06-05T01:15:29.704196+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/KWFR4RRCVD4J6GBF3FVBWRXXXS","json":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS.json","graph_json":"https://pith.science/api/pith-number/KWFR4RRCVD4J6GBF3FVBWRXXXS/graph.json","events_json":"https://pith.science/api/pith-number/KWFR4RRCVD4J6GBF3FVBWRXXXS/events.json","paper":"https://pith.science/paper/KWFR4RRC"},"agent_actions":{"view_html":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS","download_json":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS.json","view_paper":"https://pith.science/paper/KWFR4RRC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.05988&json=true","fetch_graph":"https://pith.science/api/pith-number/KWFR4RRCVD4J6GBF3FVBWRXXXS/graph.json","fetch_events":"https://pith.science/api/pith-number/KWFR4RRCVD4J6GBF3FVBWRXXXS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS/action/storage_attestation","attest_author":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS/action/author_attestation","sign_citation":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS/action/citation_signature","submit_replication":"https://pith.science/pith/KWFR4RRCVD4J6GBF3FVBWRXXXS/action/replication_record"}},"created_at":"2026-06-05T01:15:29.704196+00:00","updated_at":"2026-06-05T01:15:29.704196+00:00"}