{"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/the-hidden-life-of-tokens-reducing","title":"The Hidden Life of Tokens: Reducing Hallucination of Large Vision-Language Models via Visual Information Steering","arxiv_id":"2502.03628","date":"2025-02-05","proceeding":null,"authors":["Zhuowei Li","Haizhou Shi","Yunhe Gao","Di Liu","Zhenting Wang","Yuxiao Chen","Ting Liu","Long Zhao","Hao Wang","Dimitris N. Metaxas"],"abstract":"Large Vision-Language Models (LVLMs) can reason effectively over both textual and visual inputs, but they tend to hallucinate syntactically coherent yet visually ungrounded contents. In this paper, we investigate the internal dynamics of hallucination by examining the tokens logits rankings throughout the generation process, revealing three key patterns in how LVLMs process information: (1) gradual visual information loss -- visually grounded tokens gradually become less favored throughout generation, and (2) early excitation -- semantically meaningful tokens achieve peak activation in the layers earlier than the final layer. (3) hidden genuine information -- visually grounded tokens though not being eventually decided still retain relatively high rankings at inference. Based on these insights, we propose VISTA (Visual Information Steering with Token-logit Augmentation), a training-free inference-time intervention framework that reduces hallucination while promoting genuine information. VISTA works by combining two complementary approaches: reinforcing visual information in activation space and leveraging early layer activations to promote semantically meaningful decoding. Compared to existing methods, VISTA requires no external supervision and is applicable to various decoding strategies. Extensive experiments show that VISTA on average reduces hallucination by abount 40% on evaluated open-ended generation task, and it consistently outperforms existing methods on four benchmarks across four architectures under three decoding strategies.","url_abs":"https://arxiv.org/abs/2502.03628v1","url_pdf":"https://arxiv.org/pdf/2502.03628v1.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":"the-hidden-life-of-tokens-reducing","repo_url":"https://github.com/LzVv123456/VISTA","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"hallucination","task_name":"Hallucination"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2502.03628","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.03628"}},"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/LzVv123456/VISTA","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran_draft_wrong":2,"ran":1,"unverified":8},"by_repo_kind":{"official":{"samples":11,"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":"25585704d77cafb4","entry":"combine_coco_captions","repo":"LzVv123456/VISTA","repo_kind":"official","path":"chair_ans.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/chair_ans.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":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"25585704d77cafb4"}},{"code_sha256_prefix":"66415dc1dd50f555","entry":"combine_coco_instances","repo":"LzVv123456/VISTA","repo_kind":"official","path":"chair_ans.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/chair_ans.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":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"66415dc1dd50f555"}},{"code_sha256_prefix":"94d234b775a86dc7","entry":"get_nested_attr","repo":"LzVv123456/VISTA","repo_kind":"official","path":"llm_layers.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/llm_layers.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"94d234b775a86dc7"}},{"code_sha256_prefix":"3eeda9d3221d357b","entry":"find_longest_modulelist","repo":"LzVv123456/VISTA","repo_kind":"official","path":"llm_layers.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/llm_layers.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3eeda9d3221d357b"}},{"code_sha256_prefix":"281023289972b87d","entry":"find_module","repo":"LzVv123456/VISTA","repo_kind":"official","path":"llm_layers.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/llm_layers.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"281023289972b87d"}},{"code_sha256_prefix":"0de31df5bad47e67","entry":"init_folder_structure","repo":"LzVv123456/VISTA","repo_kind":"official","path":"myutils.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/myutils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0de31df5bad47e67"}},{"code_sha256_prefix":"bb7b96236ae47aaf","entry":"load_generated_captions","repo":"LzVv123456/VISTA","repo_kind":"official","path":"chair_ans.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/chair_ans.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"bb7b96236ae47aaf"}},{"code_sha256_prefix":"aa84cc45b40e5710","entry":"load_image","repo":"LzVv123456/VISTA","repo_kind":"official","path":"mmhal_eval.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/mmhal_eval.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"aa84cc45b40e5710"}},{"code_sha256_prefix":"7172f42b9946a5a2","entry":"load_model_args_from_yaml","repo":"LzVv123456/VISTA","repo_kind":"official","path":"model_loader.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/model_loader.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"7172f42b9946a5a2"}},{"code_sha256_prefix":"74b53a3c8567e9e6","entry":"prepare_common_fileparts","repo":"LzVv123456/VISTA","repo_kind":"official","path":"myutils.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/myutils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"74b53a3c8567e9e6"}},{"code_sha256_prefix":"3f489857c0692281","entry":"recorder","repo":"LzVv123456/VISTA","repo_kind":"official","path":"pope_ans.py","file_url":"https://github.com/LzVv123456/VISTA/blob/HEAD/pope_ans.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3f489857c0692281"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}