{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/visual-instruction-tuning-with-500x-fewer","title":"LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering","arxiv_id":"2412.12359","date":"2024-12-16","proceeding":null,"authors":["Jinhe Bi","Yujun Wang","Haokun Chen","Xun Xiao","Artur Hecker","Volker Tresp","Yunpu Ma"],"abstract":"Multimodal Large Language Models (MLLMs) have significantly advanced visual tasks by integrating visual representations into large language models (LLMs). The textual modality, inherited from LLMs, equips MLLMs with abilities like instruction following and in-context learning. In contrast, the visual modality enhances performance in downstream tasks by leveraging rich semantic content, spatial information, and grounding capabilities. These intrinsic modalities work synergistically across various visual tasks. Our research initially reveals a persistent imbalance between these modalities, with text often dominating output generation during visual instruction tuning. This imbalance occurs when using both full fine-tuning and parameter-efficient fine-tuning (PEFT) methods. We then found that re-balancing these modalities can significantly reduce the number of trainable parameters required, inspiring a direction for further optimizing visual instruction tuning. We introduce Modality Linear Representation-Steering (MoReS) to achieve the goal. MoReS effectively re-balances the intrinsic modalities throughout the model, where the key idea is to steer visual representations through linear transformations in the visual subspace across each model layer. To validate our solution, we composed LLaVA Steering, a suite of models integrated with the proposed MoReS method. Evaluation results show that the composed LLaVA Steering models require, on average, 500 times fewer trainable parameters than LoRA needs while still achieving comparable performance across three visual benchmarks and eight visual question-answering tasks. Last, we present the LLaVA Steering Factory, an in-house developed platform that enables researchers to quickly customize various MLLMs with component-based architecture for seamlessly integrating state-of-the-art models, and evaluate their intrinsic modality imbalance.","url_abs":"https://arxiv.org/abs/2412.12359v2","url_pdf":"https://arxiv.org/pdf/2412.12359v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"visual-instruction-tuning-with-500x-fewer","repo_url":"https://github.com/bibisbar/LLaVA-Steering","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"in-context-learning","task_name":"In-Context Learning"},{"task_slug":"instruction-following","task_name":"Instruction Following"},{"task_slug":"question-answering","task_name":"Question Answering"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"},{"task_slug":"parameter-efficient-fine-tuning","task_name":"parameter-efficient fine-tuning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2412.12359","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2412.12359"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/bibisbar/LLaVA-Steering","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran_draft_wrong":3,"unverified":10},"by_repo_kind":{"official":{"samples":13,"ran":3,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"42a46570620cd9fa","entry":"get_chunk","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"42a46570620cd9fa"}},{"code_sha256_prefix":"a7bee88c1c7fd6a3","entry":"image_parser","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/run_tiny_llava.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/run_tiny_llava.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a7bee88c1c7fd6a3"}},{"code_sha256_prefix":"076c252c52cbb161","entry":"split_list","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"076c252c52cbb161"}},{"code_sha256_prefix":"68ad5f668d4f2976","entry":"collate_fn","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_loader.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_loader.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"68ad5f668d4f2976"}},{"code_sha256_prefix":"3289beb8d4bdf12b","entry":"eval_model","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_halbench.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_halbench.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"3289beb8d4bdf12b"}},{"code_sha256_prefix":"b449ac23f340f344","entry":"eval_model","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_mmhal.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_mmhal.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"b449ac23f340f344"}},{"code_sha256_prefix":"65c4b77c317ba4b5","entry":"generate_questions_imgs","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_iconqa_blank.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_iconqa_blank.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"65c4b77c317ba4b5"}},{"code_sha256_prefix":"52767cf2fbfc763c","entry":"generate_questions_imgs","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_iconqa_text.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_iconqa_text.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"52767cf2fbfc763c"}},{"code_sha256_prefix":"01c6b696dad5f567","entry":"get_chunk","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_chair.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_chair.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"01c6b696dad5f567"}},{"code_sha256_prefix":"4552e857b0cf7ef7","entry":"load_image","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/run_tiny_llava.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/run_tiny_llava.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"4552e857b0cf7ef7"}},{"code_sha256_prefix":"99d66c86efac69ff","entry":"load_images","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/run_tiny_llava.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/run_tiny_llava.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"99d66c86efac69ff"}},{"code_sha256_prefix":"f3018307f9301da1","entry":"parse_multi_choice_response","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_mmmu.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_mmmu.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"f3018307f9301da1"}},{"code_sha256_prefix":"803512fe1dd82fde","entry":"split_list","repo":"bibisbar/LLaVA-Steering","repo_kind":"official","path":"tinyllava/eval/model_vqa_chair.py","file_url":"https://github.com/bibisbar/LLaVA-Steering/blob/HEAD/tinyllava/eval/model_vqa_chair.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"803512fe1dd82fde"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}