{"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/illusionvqa-a-challenging-optical-illusion","title":"IllusionVQA: A Challenging Optical Illusion Dataset for Vision Language Models","arxiv_id":"2403.15952","date":"2024-03-23","proceeding":null,"authors":["HAZ Sameen Shahgir","Khondker Salman Sayeed","Abhik Bhattacharjee","Wasi Uddin Ahmad","Yue Dong","Rifat Shahriyar"],"abstract":"The advent of Vision Language Models (VLM) has allowed researchers to investigate the visual understanding of a neural network using natural language. Beyond object classification and detection, VLMs are capable of visual comprehension and common-sense reasoning. This naturally led to the question: How do VLMs respond when the image itself is inherently unreasonable? To this end, we present IllusionVQA: a diverse dataset of challenging optical illusions and hard-to-interpret scenes to test the capability of VLMs in two distinct multiple-choice VQA tasks - comprehension and soft localization. GPT4V, the best performing VLM, achieves 62.99% accuracy (4-shot) on the comprehension task and 49.7% on the localization task (4-shot and Chain-of-Thought). Human evaluation reveals that humans achieve 91.03% and 100% accuracy in comprehension and localization. We discover that In-Context Learning (ICL) and Chain-of-Thought reasoning substantially degrade the performance of Gemini-Pro in the localization task. Tangentially, we discover a potential weakness in the ICL capabilities of VLMs: they fail to locate optical illusions even when the correct answer is in the context window as a few-shot example.","url_abs":"https://arxiv.org/abs/2403.15952v3","url_pdf":"https://arxiv.org/pdf/2403.15952v3.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":"illusionvqa-a-challenging-optical-illusion","repo_url":"https://github.com/csebuetnlp/illusionvqa","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"common-sense-reasoning","task_name":"Common Sense Reasoning"},{"task_slug":"in-context-learning","task_name":"In-Context Learning"},{"task_slug":"multiple-choice","task_name":"Multiple-choice"},{"task_slug":"object-localization","task_name":"Object Localization"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"},{"task_slug":"visual-question-answering","task_name":"Visual Question Answering (VQA)"}],"methods":[],"datasets_introduced":[{"slug":"illusionvqa","name":"IllusionVQA","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"GPT4-Vision 4-shot+CoT","rank_in_archive_order":1,"of":9,"metrics":{"Accuracy":"49.7"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"GPT4-Vision 4-shot","rank_in_archive_order":2,"of":9,"metrics":{"Accuracy":"46"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"Gemini-Pro","rank_in_archive_order":3,"of":9,"metrics":{"Accuracy":"43.5"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"Gemini-Pro 4-shot","rank_in_archive_order":4,"of":9,"metrics":{"Accuracy":"41.8"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"GPT4-Vision","rank_in_archive_order":5,"of":9,"metrics":{"Accuracy":"40"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"Gemini-Pro 4-shot+CoT","rank_in_archive_order":6,"of":9,"metrics":{"Accuracy":"33.9"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"CogVLM","rank_in_archive_order":7,"of":9,"metrics":{"Accuracy":"28"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"LLaVA-1.5-13B","rank_in_archive_order":8,"of":9,"metrics":{"Accuracy":"24.8"},"uses_additional_data":false},{"leaderboard":"/sota/object-localization-on-illusionvqa","task":"Object Localization","dataset":"IllusionVQA","model":"InstructBLIP-13B","rank_in_archive_order":9,"of":9,"metrics":{"Accuracy":"24.3"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"GPT4-Vision 4-shot","rank_in_archive_order":1,"of":7,"metrics":{"Accuracy":"62.99"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"GPT4-Vision","rank_in_archive_order":2,"of":7,"metrics":{"Accuracy":"58.85"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"Gemini-Pro 4-shot","rank_in_archive_order":3,"of":7,"metrics":{"Accuracy":"52.87"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"Gemini-Pro","rank_in_archive_order":4,"of":7,"metrics":{"Accuracy":"51.26"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"LLaVA-1.5-13B","rank_in_archive_order":5,"of":7,"metrics":{"Accuracy":"40"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"CogVLM","rank_in_archive_order":6,"of":7,"metrics":{"Accuracy":"38.16"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-vqa-on-illusionvqa","task":"Visual Question Answering (VQA)","dataset":"IllusionVQA","model":"InstructBLIP-13B","rank_in_archive_order":7,"of":7,"metrics":{"Accuracy":"34.25"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2403.15952","atlas_url":"https://app.syntology.ai/?focus=2403.15952","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2403.15952"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+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/csebuetnlp/illusionvqa","reach":{"status":"ok"}}],"summary":{"ran":3,"ran_draft_wrong":1,"ran_fixture":1},"by_repo_kind":{"official":{"samples":5,"ran":5,"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":5,"samples":[{"code_sha256_prefix":"57e95166ade5f25c","entry":"build_transform","repo":"csebuetnlp/illusionvqa","repo_kind":"official","path":"inference_code/open_source/internvlm_inference.py","file_url":"https://github.com/csebuetnlp/illusionvqa/blob/HEAD/inference_code/open_source/internvlm_inference.py","link_basis":"harvester_set","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":"57e95166ade5f25c"}},{"code_sha256_prefix":"0a3d998193bf71de","entry":"construct_mcq","repo":"csebuetnlp/illusionvqa","repo_kind":"official","path":"inference_code/open_source/cogvlm2_inference.py","file_url":"https://github.com/csebuetnlp/illusionvqa/blob/HEAD/inference_code/open_source/cogvlm2_inference.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":"0a3d998193bf71de"}},{"code_sha256_prefix":"c2ec19003bda7600","entry":"construct_mcq","repo":"csebuetnlp/illusionvqa","repo_kind":"official","path":"inference_code/open_source/instructblip_inference.py","file_url":"https://github.com/csebuetnlp/illusionvqa/blob/HEAD/inference_code/open_source/instructblip_inference.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":"c2ec19003bda7600"}},{"code_sha256_prefix":"6dc77bb22fb9d660","entry":"dynamic_preprocess","repo":"csebuetnlp/illusionvqa","repo_kind":"official","path":"inference_code/open_source/internvlm_inference.py","file_url":"https://github.com/csebuetnlp/illusionvqa/blob/HEAD/inference_code/open_source/internvlm_inference.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"6dc77bb22fb9d660"}},{"code_sha256_prefix":"bd77f5f8067f18e9","entry":"find_closest_aspect_ratio","repo":"csebuetnlp/illusionvqa","repo_kind":"official","path":"inference_code/open_source/internvlm_inference.py","file_url":"https://github.com/csebuetnlp/illusionvqa/blob/HEAD/inference_code/open_source/internvlm_inference.py","link_basis":"harvester_set","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"bd77f5f8067f18e9"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}