{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:7BAOA4VSGL4XGG2I5XHN3HKGQM","short_pith_number":"pith:7BAOA4VS","schema_version":"1.0","canonical_sha256":"f840e072b232f9731b48edcedd9d46830cdd56eb528d1784267a2cbfd99accd9","source":{"kind":"arxiv","id":"2407.07093","version":1},"attestation_state":"computed","paper":{"title":"FBI-LLM: Scaling Up Fully Binarized LLMs from Scratch via Autoregressive Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Liqun Ma, Mingjie Sun, Zhiqiang Shen","submitted_at":"2024-07-09T17:59:48Z","abstract_excerpt":"This work presents a Fully BInarized Large Language Model (FBI-LLM), demonstrating for the first time how to train a large-scale binary language model from scratch (not the partial binary or ternary LLM like BitNet b1.58) to match the performance of its full-precision counterparts (e.g., FP16 or BF16) in transformer-based LLMs. It achieves this by employing an autoregressive distillation (AD) loss with maintaining equivalent model dimensions (130M, 1.3B, 7B) and training data volume as regular LLM pretraining, while delivering competitive results in terms of perplexity and task-specific effect"},"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":"2407.07093","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-09T17:59:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"629e57c6bec53e2a45beba25362b3ffeb550fe2354b79038a6dd6976bc64112e","abstract_canon_sha256":"c7f055d78c25789bd0293e39c5803f1ac99bc457e7deca7c3379a70370db07f4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:41:56.831024Z","signature_b64":"wVNh62u1uZPJD1AwVi3fgNS99CmUsQ4qJkdg8U+nVp4yh/RQG7vQcjCnAaqv/2Mrlx61/UZ+Fv2SKtdkjZYcDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f840e072b232f9731b48edcedd9d46830cdd56eb528d1784267a2cbfd99accd9","last_reissued_at":"2026-07-05T08:41:56.830615Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:41:56.830615Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FBI-LLM: Scaling Up Fully Binarized LLMs from Scratch via Autoregressive Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Liqun Ma, Mingjie Sun, Zhiqiang Shen","submitted_at":"2024-07-09T17:59:48Z","abstract_excerpt":"This work presents a Fully BInarized Large Language Model (FBI-LLM), demonstrating for the first time how to train a large-scale binary language model from scratch (not the partial binary or ternary LLM like BitNet b1.58) to match the performance of its full-precision counterparts (e.g., FP16 or BF16) in transformer-based LLMs. It achieves this by employing an autoregressive distillation (AD) loss with maintaining equivalent model dimensions (130M, 1.3B, 7B) and training data volume as regular LLM pretraining, while delivering competitive results in terms of perplexity and task-specific effect"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.07093","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/2407.07093/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":"2407.07093","created_at":"2026-07-05T08:41:56.830672+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.07093v1","created_at":"2026-07-05T08:41:56.830672+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.07093","created_at":"2026-07-05T08:41:56.830672+00:00"},{"alias_kind":"pith_short_12","alias_value":"7BAOA4VSGL4X","created_at":"2026-07-05T08:41:56.830672+00:00"},{"alias_kind":"pith_short_16","alias_value":"7BAOA4VSGL4XGG2I","created_at":"2026-07-05T08:41:56.830672+00:00"},{"alias_kind":"pith_short_8","alias_value":"7BAOA4VS","created_at":"2026-07-05T08:41:56.830672+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2508.06974","citing_title":"Rethinking 1-bit Optimization Leveraging Pre-trained Large Language Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2506.12040","citing_title":"BTC-LLM: Efficient Sub-1-Bit LLM Quantization via Learnable Transformation and Binary Codebook","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19167","citing_title":"LBLLM: Lightweight Binarization of Large Language Models via Three-Stage Distillation","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM","json":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM.json","graph_json":"https://pith.science/api/pith-number/7BAOA4VSGL4XGG2I5XHN3HKGQM/graph.json","events_json":"https://pith.science/api/pith-number/7BAOA4VSGL4XGG2I5XHN3HKGQM/events.json","paper":"https://pith.science/paper/7BAOA4VS"},"agent_actions":{"view_html":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM","download_json":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM.json","view_paper":"https://pith.science/paper/7BAOA4VS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.07093&json=true","fetch_graph":"https://pith.science/api/pith-number/7BAOA4VSGL4XGG2I5XHN3HKGQM/graph.json","fetch_events":"https://pith.science/api/pith-number/7BAOA4VSGL4XGG2I5XHN3HKGQM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM/action/storage_attestation","attest_author":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM/action/author_attestation","sign_citation":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM/action/citation_signature","submit_replication":"https://pith.science/pith/7BAOA4VSGL4XGG2I5XHN3HKGQM/action/replication_record"}},"created_at":"2026-07-05T08:41:56.830672+00:00","updated_at":"2026-07-05T08:41:56.830672+00:00"}