{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NMYYVRRQRMBNXT2MGI7T6LC2FP","short_pith_number":"pith:NMYYVRRQ","schema_version":"1.0","canonical_sha256":"6b318ac6308b02dbcf4c323f3f2c5a2be1ef5f0ad9d59d0cb77eaeb17f3561ce","source":{"kind":"arxiv","id":"2401.00788","version":1},"attestation_state":"computed","paper":{"title":"Astraios: Parameter-Efficient Instruction Tuning Code Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.SE"],"primary_cat":"cs.CL","authors_text":"Armel Zebaze, Harm de Vries, Leandro Von Werra, Niklas Muennighoff, Nitchakarn Suppattarachai, Qian Liu, Terry Yue Zhuo","submitted_at":"2024-01-01T15:30:19Z","abstract_excerpt":"The high cost of full-parameter fine-tuning (FFT) of Large Language Models (LLMs) has led to a series of parameter-efficient fine-tuning (PEFT) methods. However, it remains unclear which methods provide the best cost-performance trade-off at different model scales. We introduce Astraios, a suite of 28 instruction-tuned OctoCoder models using 7 tuning methods and 4 model sizes up to 16 billion parameters. Through investigations across 5 tasks and 8 different datasets encompassing both code comprehension and code generation tasks, we find that FFT generally leads to the best downstream performan"},"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":"2401.00788","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-01-01T15:30:19Z","cross_cats_sorted":["cs.AI","cs.SE"],"title_canon_sha256":"49c76e06784c83f7f8cff4ce7dd117b2ecf033548a2410008a9ba017e45542d4","abstract_canon_sha256":"49ca2d1ec3e0406e6f971548bd2996fe57877bff688af63fc1939f8a1197d9fc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:29:22.856296Z","signature_b64":"0Vm2bd2200Ja0o1m/whts+9Qb0UaPVjTS6p0y2xNoGZq2ED769ysPVhGOSafjMj3fig3wxz7TYhHp34HYG+yAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6b318ac6308b02dbcf4c323f3f2c5a2be1ef5f0ad9d59d0cb77eaeb17f3561ce","last_reissued_at":"2026-07-05T07:29:22.855812Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:29:22.855812Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Astraios: Parameter-Efficient Instruction Tuning Code Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.SE"],"primary_cat":"cs.CL","authors_text":"Armel Zebaze, Harm de Vries, Leandro Von Werra, Niklas Muennighoff, Nitchakarn Suppattarachai, Qian Liu, Terry Yue Zhuo","submitted_at":"2024-01-01T15:30:19Z","abstract_excerpt":"The high cost of full-parameter fine-tuning (FFT) of Large Language Models (LLMs) has led to a series of parameter-efficient fine-tuning (PEFT) methods. However, it remains unclear which methods provide the best cost-performance trade-off at different model scales. We introduce Astraios, a suite of 28 instruction-tuned OctoCoder models using 7 tuning methods and 4 model sizes up to 16 billion parameters. Through investigations across 5 tasks and 8 different datasets encompassing both code comprehension and code generation tasks, we find that FFT generally leads to the best downstream performan"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.00788","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/2401.00788/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":"2401.00788","created_at":"2026-07-05T07:29:22.855875+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.00788v1","created_at":"2026-07-05T07:29:22.855875+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.00788","created_at":"2026-07-05T07:29:22.855875+00:00"},{"alias_kind":"pith_short_12","alias_value":"NMYYVRRQRMBN","created_at":"2026-07-05T07:29:22.855875+00:00"},{"alias_kind":"pith_short_16","alias_value":"NMYYVRRQRMBNXT2M","created_at":"2026-07-05T07:29:22.855875+00:00"},{"alias_kind":"pith_short_8","alias_value":"NMYYVRRQ","created_at":"2026-07-05T07:29:22.855875+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07619","citing_title":"Rethinking Code Performance Benchmarks for LLMs","ref_index":83,"is_internal_anchor":true},{"citing_arxiv_id":"2605.19018","citing_title":"LoRA vs. Full Fine-Tuning: A Theoretical Perspective","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09134","citing_title":"BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09134","citing_title":"BoostAPR: Boosting Automated Program Repair via Execution-Grounded Reinforcement Learning with Dual Reward Models","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP","json":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP.json","graph_json":"https://pith.science/api/pith-number/NMYYVRRQRMBNXT2MGI7T6LC2FP/graph.json","events_json":"https://pith.science/api/pith-number/NMYYVRRQRMBNXT2MGI7T6LC2FP/events.json","paper":"https://pith.science/paper/NMYYVRRQ"},"agent_actions":{"view_html":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP","download_json":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP.json","view_paper":"https://pith.science/paper/NMYYVRRQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.00788&json=true","fetch_graph":"https://pith.science/api/pith-number/NMYYVRRQRMBNXT2MGI7T6LC2FP/graph.json","fetch_events":"https://pith.science/api/pith-number/NMYYVRRQRMBNXT2MGI7T6LC2FP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP/action/storage_attestation","attest_author":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP/action/author_attestation","sign_citation":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP/action/citation_signature","submit_replication":"https://pith.science/pith/NMYYVRRQRMBNXT2MGI7T6LC2FP/action/replication_record"}},"created_at":"2026-07-05T07:29:22.855875+00:00","updated_at":"2026-07-05T07:29:22.855875+00:00"}