{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:KW54KIDFZCTH4FO4O4RS6VPIV7","short_pith_number":"pith:KW54KIDF","schema_version":"1.0","canonical_sha256":"55bbc52065c8a67e15dc77232f55e8afc43b5a16ab565fd34a0b47466d3502a0","source":{"kind":"arxiv","id":"2303.16199","version":3},"attestation_state":"computed","paper":{"title":"LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LLaMA-Adapter adapts frozen LLaMA to follow instructions using only 1.2 million added parameters.","cross_cats":["cs.AI","cs.CL","cs.LG","cs.MM"],"primary_cat":"cs.CV","authors_text":"Aojun Zhou, Chris Liu, Hongsheng Li, Jiaming Han, Pan Lu, Peng Gao, Renrui Zhang, Shilin Yan, Xiangfei Hu, Yu Qiao","submitted_at":"2023-03-28T17:59:12Z","abstract_excerpt":"We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively pres"},"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":true,"formal_links_present":false},"canonical_record":{"source":{"id":"2303.16199","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-03-28T17:59:12Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG","cs.MM"],"title_canon_sha256":"831ffdfc694e4500642984da60782b2e5f1362bd1a2f56a85198952d10fff66a","abstract_canon_sha256":"93d3c5dddc0333e446f23e0119f911c68110fb647f6e9d9a5b4dc35c69d290b1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:19.918526Z","signature_b64":"+IWjWxXZ/fywm56IzaBdXj8oSF1oOvrIxY0G+5gAL6FyBmmw4l+yeKQCqbLC1pNtuoGfw5kwXnTUfhAtbSbJAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"55bbc52065c8a67e15dc77232f55e8afc43b5a16ab565fd34a0b47466d3502a0","last_reissued_at":"2026-05-17T23:39:19.917730Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:19.917730Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LLaMA-Adapter adapts frozen LLaMA to follow instructions using only 1.2 million added parameters.","cross_cats":["cs.AI","cs.CL","cs.LG","cs.MM"],"primary_cat":"cs.CV","authors_text":"Aojun Zhou, Chris Liu, Hongsheng Li, Jiaming Han, Pan Lu, Peng Gao, Renrui Zhang, Shilin Yan, Xiangfei Hu, Yu Qiao","submitted_at":"2023-03-28T17:59:12Z","abstract_excerpt":"We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively pres"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The zero-initialized attention mechanism with zero gating adaptively injects the new instructional cues into LLaMA while effectively preserving its pre-trained knowledge.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LLaMA-Adapter adapts frozen LLaMA to follow instructions using only 1.2 million added parameters.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"50a8db6213e450be59e73c21dc566ff19f998250c31b89e53264e3b522e0b4ee"},"source":{"id":"2303.16199","kind":"arxiv","version":3},"verdict":{"id":"033679df-4a70-4f12-9af8-1266f199bfda","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T23:01:25.060113Z","strongest_claim":"With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters.","one_line_summary":"LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The zero-initialized attention mechanism with zero gating adaptively injects the new instructional cues into LLaMA while effectively preserving its pre-trained knowledge.","pith_extraction_headline":"LLaMA-Adapter adapts frozen LLaMA to follow instructions using only 1.2 million added parameters."},"references":{"count":278,"sample":[{"doi":"","year":2023,"title":"Alpaca-lora. https://github.com/tloen/alpaca-lora, 2023","work_id":"9a06075c-ae6a-4dd1-a289-ebde72983a0f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Flamingo: a visual language model for few-shot learning","work_id":"01d90a48-13d8-4bc8-a06a-97adb5201146","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Open llm leaderboard","work_id":"5d0d6bb6-6bf7-45f4-a39e-40222560248f","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1901,"title":"Language models are few-shot learners","work_id":"b5af3a68-2622-4421-b39b-b1d2fbde2d8d","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2004,"title":"Introduction to the conll-2004 shared task: Semantic role labeling","work_id":"c0bd2045-d243-4c6d-a996-2634541c2b6b","ref_index":6,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":278,"snapshot_sha256":"14c936933f9c610e90f0dbbeae5ab392bfc5944dc15cd86e833a0b401fe6c09b","internal_anchors":45},"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":"2303.16199","created_at":"2026-05-17T23:39:19.917874+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.16199v3","created_at":"2026-05-17T23:39:19.917874+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.16199","created_at":"2026-05-17T23:39:19.917874+00:00"},{"alias_kind":"pith_short_12","alias_value":"KW54KIDFZCTH","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"KW54KIDFZCTH4FO4","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"KW54KIDF","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":44,"internal_anchor_count":23,"sample":[{"citing_arxiv_id":"2306.00978","citing_title":"AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2308.12067","citing_title":"MM-LIMA: Less Is More for Alignment in Multi-Modal Datasets","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2407.08101","citing_title":"What to Say and When to Say it: Live Fitness Coaching as a Testbed for Situated Interaction","ref_index":66,"is_internal_anchor":true},{"citing_arxiv_id":"2501.05067","citing_title":"LLaVA-Octopus: Unlocking Instruction-Driven Adaptive Projector Fusion for Video Understanding","ref_index":89,"is_internal_anchor":true},{"citing_arxiv_id":"2503.16549","citing_title":"MathFlow: Enhancing the Perceptual Flow of MLLMs for Visual Mathematical Problems","ref_index":78,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18678","citing_title":"Lance: Unified Multimodal Modeling by Multi-Task Synergy","ref_index":147,"is_internal_anchor":true},{"citing_arxiv_id":"2605.18678","citing_title":"Lance: Unified Multimodal Modeling by Multi-Task Synergy","ref_index":146,"is_internal_anchor":true},{"citing_arxiv_id":"2307.06435","citing_title":"A Comprehensive Overview of Large Language Models","ref_index":285,"is_internal_anchor":true},{"citing_arxiv_id":"2506.18962","citing_title":"UniMind: Unleashing the Power of LLMs for Unified Multi-Task Brain Decoding","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2509.03740","citing_title":"CLIP-SVD: Efficient and Interpretable Vision-Language Adaptation via Singular Values","ref_index":79,"is_internal_anchor":true},{"citing_arxiv_id":"2311.17005","citing_title":"MVBench: A Comprehensive Multi-modal Video Understanding Benchmark","ref_index":103,"is_internal_anchor":true},{"citing_arxiv_id":"2312.13771","citing_title":"AppAgent: Multimodal Agents as Smartphone Users","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2502.04326","citing_title":"WorldSense: Evaluating Real-world Omnimodal Understanding for Multimodal LLMs","ref_index":83,"is_internal_anchor":true},{"citing_arxiv_id":"2512.02764","citing_title":"PEFT-Factory: Unified Parameter-Efficient Fine-Tuning of Autoregressive Large Language Models","ref_index":81,"is_internal_anchor":true},{"citing_arxiv_id":"2407.01284","citing_title":"We-Math: Does Your Large Multimodal Model Achieve Human-like Mathematical Reasoning?","ref_index":11,"is_internal_anchor":true},{"citing_arxiv_id":"2402.08268","citing_title":"World Model on Million-Length Video And Language With Blockwise RingAttention","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2306.13549","citing_title":"A Survey on Multimodal Large Language Models","ref_index":78,"is_internal_anchor":true},{"citing_arxiv_id":"2503.10631","citing_title":"HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Model","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2404.16994","citing_title":"PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning","ref_index":51,"is_internal_anchor":true},{"citing_arxiv_id":"2304.15010","citing_title":"LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model","ref_index":72,"is_internal_anchor":true},{"citing_arxiv_id":"2311.16502","citing_title":"MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI","ref_index":89,"is_internal_anchor":true},{"citing_arxiv_id":"2305.03726","citing_title":"Otter: A Multi-Modal Model with In-Context Instruction Tuning","ref_index":100,"is_internal_anchor":true},{"citing_arxiv_id":"2303.16199","citing_title":"LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention","ref_index":282,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10819","citing_title":"ALAM: Algebraically Consistent Latent Action Model for Vision-Language-Action Models","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2311.10122","citing_title":"Video-LLaVA: Learning United Visual Representation by Alignment Before Projection","ref_index":81,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7","json":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7.json","graph_json":"https://pith.science/api/pith-number/KW54KIDFZCTH4FO4O4RS6VPIV7/graph.json","events_json":"https://pith.science/api/pith-number/KW54KIDFZCTH4FO4O4RS6VPIV7/events.json","paper":"https://pith.science/paper/KW54KIDF"},"agent_actions":{"view_html":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7","download_json":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7.json","view_paper":"https://pith.science/paper/KW54KIDF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.16199&json=true","fetch_graph":"https://pith.science/api/pith-number/KW54KIDFZCTH4FO4O4RS6VPIV7/graph.json","fetch_events":"https://pith.science/api/pith-number/KW54KIDFZCTH4FO4O4RS6VPIV7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7/action/storage_attestation","attest_author":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7/action/author_attestation","sign_citation":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7/action/citation_signature","submit_replication":"https://pith.science/pith/KW54KIDFZCTH4FO4O4RS6VPIV7/action/replication_record"}},"created_at":"2026-05-17T23:39:19.917874+00:00","updated_at":"2026-05-17T23:39:19.917874+00:00"}