{"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/eta-evaluating-then-aligning-safety-of-vision","title":"ETA: Evaluating Then Aligning Safety of Vision Language Models at Inference Time","arxiv_id":"2410.06625","date":"2024-10-09","proceeding":null,"authors":["Yi Ding","Bolian Li","Ruqi Zhang"],"abstract":"Vision Language Models (VLMs) have become essential backbones for multimodal intelligence, yet significant safety challenges limit their real-world application. While textual inputs are often effectively safeguarded, adversarial visual inputs can easily bypass VLM defense mechanisms. Existing defense methods are either resource-intensive, requiring substantial data and compute, or fail to simultaneously ensure safety and usefulness in responses. To address these limitations, we propose a novel two-phase inference-time alignment framework, Evaluating Then Aligning (ETA): 1) Evaluating input visual contents and output responses to establish a robust safety awareness in multimodal settings, and 2) Aligning unsafe behaviors at both shallow and deep levels by conditioning the VLMs' generative distribution with an interference prefix and performing sentence-level best-of-N to search the most harmless and helpful generation paths. Extensive experiments show that ETA outperforms baseline methods in terms of harmlessness, helpfulness, and efficiency, reducing the unsafe rate by 87.5% in cross-modality attacks and achieving 96.6% win-ties in GPT-4 helpfulness evaluation. The code is publicly available at https://github.com/DripNowhy/ETA.","url_abs":"https://arxiv.org/abs/2410.06625v1","url_pdf":"https://arxiv.org/pdf/2410.06625v1.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":"eta-evaluating-then-aligning-safety-of-vision","repo_url":"https://github.com/dripnowhy/eta","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"sentence","task_name":"Sentence"}],"methods":[{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gpt-4","method_name":"GPT-4"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2410.06625","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.06625"}},"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/dripnowhy/eta","reach":{"status":"ok"}}],"summary":{"ran":5,"ran_draft_wrong":3,"ran_violates":1,"unverified":2},"by_repo_kind":{"official":{"samples":11,"ran":9,"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":11,"samples":[{"code_sha256_prefix":"da2114a8052b6d55","entry":"calculate_safety_scores","repo":"dripnowhy/eta","repo_kind":"official","path":"eval_usr.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/eval_usr.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"da2114a8052b6d55"}},{"code_sha256_prefix":"8e8f6b744fcf648c","entry":"check_success_attack","repo":"dripnowhy/eta","repo_kind":"official","path":"eval_asr.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/eval_asr.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"8e8f6b744fcf648c"}},{"code_sha256_prefix":"b3b9db8ec8e0d1d1","entry":"collate_fn","repo":"dripnowhy/eta","repo_kind":"official","path":"llava/eval/model_vqa_loader_eta.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/llava/eval/model_vqa_loader_eta.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"b3b9db8ec8e0d1d1"}},{"code_sha256_prefix":"f41cb1a19b154297","entry":"encode_image","repo":"dripnowhy/eta","repo_kind":"official","path":"helpfulscore_ai.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/helpfulscore_ai.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":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"f41cb1a19b154297"}},{"code_sha256_prefix":"42a46570620cd9fa","entry":"get_chunk","repo":"dripnowhy/eta","repo_kind":"official","path":"llava/eval/model_vqa_loader_eta.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/llava/eval/model_vqa_loader_eta.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":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"42a46570620cd9fa"}},{"code_sha256_prefix":"eae64eb562f6b0d1","entry":"gpt4_eval","repo":"dripnowhy/eta","repo_kind":"official","path":"helpfulscore_ai.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/helpfulscore_ai.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"eae64eb562f6b0d1"}},{"code_sha256_prefix":"6d82575864d8ddeb","entry":"gpt4_eval","repo":"dripnowhy/eta","repo_kind":"official","path":"helpful_wintie_ai.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/helpful_wintie_ai.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"6d82575864d8ddeb"}},{"code_sha256_prefix":"bae18947b56f2be1","entry":"is_none","repo":"dripnowhy/eta","repo_kind":"official","path":"llava/eval/model_vqa_mmbench_eta.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/llava/eval/model_vqa_mmbench_eta.py","link_basis":"harvester_set","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"bae18947b56f2be1"}},{"code_sha256_prefix":"076c252c52cbb161","entry":"split_list","repo":"dripnowhy/eta","repo_kind":"official","path":"llava/eval/model_vqa_loader_eta.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/llava/eval/model_vqa_loader_eta.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"076c252c52cbb161"}},{"code_sha256_prefix":"85e3fe4d95b795f8","entry":"evaluate_safety","repo":"dripnowhy/eta","repo_kind":"official","path":"eval_usr.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/eval_usr.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"85e3fe4d95b795f8"}},{"code_sha256_prefix":"14c54d6727f29560","entry":"get_stopping_criteria","repo":"dripnowhy/eta","repo_kind":"official","path":"modeling_internlm_xcomposer2.py","file_url":"https://github.com/dripnowhy/eta/blob/HEAD/modeling_internlm_xcomposer2.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"14c54d6727f29560"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}