{"id":20046630,"url":"https://github.com/ncsoft/idk","last_synced_at":"2026-03-05T19:50:47.718Z","repository":{"id":222594319,"uuid":"757189858","full_name":"ncsoft/idk","owner":"ncsoft","description":"Official implementation of \"Visually Dehallucinative Instruction Generation: Know What You Don't Know\"","archived":false,"fork":false,"pushed_at":"2024-02-16T06:36:28.000Z","size":4513,"stargazers_count":6,"open_issues_count":0,"forks_count":2,"subscribers_count":4,"default_branch":"main","last_synced_at":"2025-01-12T20:12:44.702Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":"","language":null,"has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"bsd-3-clause","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/ncsoft.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-02-14T01:08:04.000Z","updated_at":"2024-09-19T06:56:16.000Z","dependencies_parsed_at":"2024-02-15T06:25:51.184Z","dependency_job_id":"405573c3-6ea9-46eb-8fcb-5935d3e942d9","html_url":"https://github.com/ncsoft/idk","commit_stats":null,"previous_names":["ncsoft/idk"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ncsoft%2Fidk","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ncsoft%2Fidk/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ncsoft%2Fidk/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/ncsoft%2Fidk/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/ncsoft","download_url":"https://codeload.github.com/ncsoft/idk/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":241476435,"owners_count":19968916,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-11-13T11:25:19.609Z","updated_at":"2026-03-05T19:50:42.679Z","avatar_url":"https://github.com/ncsoft.png","language":null,"funding_links":[],"categories":[],"sub_categories":[],"readme":"# Visually Dehallucinative Instruction Generation: Know What You Don't Know\n\n**Visually Dehallucinative Instruction Generation: Know What You Don't Know** [[paper](https://arxiv.org/abs/2402.09717)] \u003cbr\u003e\n[Sungguk Cha](https://sunggukcha.github.io), Jusung Lee, Younghyun Lee and Cheoljong Yang\n\n\u003cimg src=\"images/categories.png\"\u003e \u003cbr\u003e\n\nThis work proposes ***I Know Hallucination*** wherein hallucination-inducing question comes and the desired answer is \"I Don't Know.\"\n\n## IDK-Instructions\n\u003cimg src=\"images/qualitative.png\"\u003e \u003cbr\u003e\nQualitative results on hallucination inducing-questions. \n\nPrepare COCO images in advance. \n\nIDK-Instructions uses \"\\\u003cimage\\\u003e\" for the image token.\n```\n{\n        \"image\": \"MSCOCO/images/train2014/COCO_train2014_000000393224.jpg\",\n        \"conversations\": [\n            {\n                \"from\": \"human\",\n                \"value\": \"\u003cimage\u003eWhat is the business man doing in the picture?\"\n            },\n            {\n                \"from\": \"gpt\",\n                \"value\": \"It is unknown what the business man is doing in the picture. It could be waiting, standing, walking or frowning.\"\n            },\n            {\n                \"from\": \"human\",\n                \"value\": \"What is the business man doing in the picture?\"\n            },\n            {\n                \"from\": \"gpt\",\n                \"value\": \"It is unclear what the business man is doing in the picture. He could be waiting, standing, walking, or even frowning or pouting.\"\n            }\n        ]\n    },\n```\n\nIDK-Instructions training split comprises 13,807 questions with a total of 27,614 answers, while the validation split consists of 6,624 questions with a total of 13,248 answers.\nPiling question-answer pairs with respect to the same image, resulting dialogue formed instruction has 11,123 and 5,496 dialogues for train and validation splits, respectively.\n\n## VQAv2-IDK\n\u003cimg src=\"images/vqav2idk.png\"\u003e \u003cbr\u003e\n\nVQAv2-IDK is the subset of VQAv2 dataset, consisting of unanswerable (in other words, hallucination-inducing) image-questions, where the desired answer becomes \"I Don't Know\".\n\n## Citation\nIf you find it useful for your research and applications, please cite using this BibTeX:\n```\n@inproceedings{cha2024visually,\n      title={Visually Dehallucinative Instruction Generation: Know What You Don't Know}, \n      author={Cha, Sungguk and Lee, Jusung and Lee, Younghyun and Yang, Cheoljong},\n      year={2024},\n}\n```\n\n## Licenses\nThis work used VQAv2 dataset (CC BY 4.0 DEED license) for the question-answer source and ChatGPT for IDK-Instructions generation (refer OpenAI policies, https://openai.com/policies).\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fncsoft%2Fidk","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fncsoft%2Fidk","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fncsoft%2Fidk/lists"}