{"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/rlaif-v-aligning-mllms-through-open-source-ai","title":"RLAIF-V: Open-Source AI Feedback Leads to Super GPT-4V Trustworthiness","arxiv_id":"2405.17220","date":"2024-05-27","proceeding":"CVPR 2025 1","authors":["Tianyu Yu","Haoye Zhang","Qiming Li","Qixin Xu","Yuan YAO","Da Chen","Xiaoman Lu","Ganqu Cui","Yunkai Dang","Taiwen He","Xiaocheng Feng","Jun Song","Bo Zheng","Zhiyuan Liu","Tat-Seng Chua","Maosong Sun"],"abstract":"Traditional feedback learning for hallucination reduction relies on labor-intensive manual labeling or expensive proprietary models. This leaves the community without foundational knowledge about how to build high-quality feedback with open-source MLLMs. In this work, we introduce RLAIF-V, a novel framework that aligns MLLMs in a fully open-source paradigm. RLAIF-V maximally explores open-source MLLMs from two perspectives, including high-quality feedback data generation for preference learning and self-feedback guidance for inference-time scaling. Extensive experiments on six benchmarks in both automatic and human evaluation show that RLAIF-V substantially enhances the trustworthiness of models at both preference learning and inference time. RLAIF-V 7B reduces object hallucination by 80.7\\% and overall hallucination by 33.7\\%. Remarkably, RLAIF-V 12B further reveals the self-alignment potential of open-source MLLMs, where the model can learn from feedback of itself to achieve super GPT-4V trustworthiness.","url_abs":"https://arxiv.org/abs/2405.17220v2","url_pdf":"https://arxiv.org/pdf/2405.17220v2.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":"rlaif-v-aligning-mllms-through-open-source-ai","repo_url":"https://github.com/openbmb/omnilmm","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"rlaif-v-aligning-mllms-through-open-source-ai","repo_url":"https://github.com/rlhf-v/rlaif-v","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"rlaif-v-aligning-mllms-through-open-source-ai","repo_url":"https://github.com/OpenBMB/MiniCPM-o","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"rlaif-v-aligning-mllms-through-open-source-ai","repo_url":"https://github.com/openbmb/minicpm-v","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"rlaif-v-aligning-mllms-through-open-source-ai","repo_url":"https://github.com/rlhf-v/rlhf-v","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}}],"tasks":[{"task_slug":"hallucination","task_name":"Hallucination"},{"task_slug":"image-captioning","task_name":"Image Captioning"},{"task_slug":"object-hallucination","task_name":"Object Hallucination"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"}],"methods":[],"datasets_introduced":[{"slug":"rlaif-v-dataset","name":"RLAIF-V Dataset","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-captioning-on-object-halbench","task":"Image Captioning","dataset":"Object HalBench","model":"RLAIF-V 7B","rank_in_archive_order":2,"of":3,"metrics":{"chair_i":"4.3","chair_s":"8.5"},"uses_additional_data":false},{"leaderboard":"/sota/image-captioning-on-object-halbench","task":"Image Captioning","dataset":"Object HalBench","model":"RLAIF-V 12B","rank_in_archive_order":3,"of":3,"metrics":{"chair_i":"1.8","chair_s":"3.3"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-amber","task":"Visual Question Answering","dataset":"AMBER","model":"RLAIF-V 12B","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"88","F1":"90.9"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-mmhal-bench","task":"Visual Question Answering","dataset":"MMHal-Bench","model":"RLAIF-V 12B","rank_in_archive_order":1,"of":2,"metrics":{"Hallucination Rate":"29.2","Score":"3.36"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-mmhal-bench","task":"Visual Question Answering","dataset":"MMHal-Bench","model":"RLAIF-V 7B","rank_in_archive_order":2,"of":2,"metrics":{"Hallucination Rate":"29.2","Score":"3.06"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2405.17220","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2405.17220"}},"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/rlhf-v/rlaif-v","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/OpenBMB/MiniCPM-o","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/openbmb/minicpm-v","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/openbmb/omnilmm","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/rlhf-v/rlhf-v","reach":{"status":"ok"}}],"summary":{"ran_fixture":3,"ran":3,"ran_draft_wrong":6,"ran_violates":2,"unverified":8},"by_repo_kind":{"official":{"samples":15,"ran":8,"repositories":2},"listed":{"samples":7,"ran":6,"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":8,"samples":[{"code_sha256_prefix":"a95e6290546a4d45","entry":"autocontrast_func","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/model/utils.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/model/utils.py","link_basis":"plan_row","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"DEP_MISSING","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a95e6290546a4d45"}},{"code_sha256_prefix":"03a2b1b7f5fd092e","entry":"cal_informative","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/summarize_gpt_mmhal_review.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/summarize_gpt_mmhal_review.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":"03a2b1b7f5fd092e"}},{"code_sha256_prefix":"25585704d77cafb4","entry":"combine_coco_captions","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/eval_gpt_obj_halbench.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/eval_gpt_obj_halbench.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":"25585704d77cafb4"}},{"code_sha256_prefix":"eb4f4c5587a11a2c","entry":"equalize_func","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/model/utils.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/model/utils.py","link_basis":"plan_row","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"DEP_MISSING","metamorphic_tier":"invariant","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"eb4f4c5587a11a2c"}},{"code_sha256_prefix":"f067ba49686d26ba","entry":"expand_question_into_multimodal","repo":"rlhf-v/rlaif-v","repo_kind":"official","path":"chat.py","file_url":"https://github.com/rlhf-v/rlaif-v/blob/HEAD/chat.py","link_basis":"first_harvest_node","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":"f067ba49686d26ba"}},{"code_sha256_prefix":"77e8a3ac46f3afec","entry":"get_2d_sincos_pos_embed","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/model/resampler.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/model/resampler.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"77e8a3ac46f3afec"}},{"code_sha256_prefix":"50c4398a8645438a","entry":"get_abs_pos","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/model/resampler.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/model/resampler.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"MISDECLARED","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"50c4398a8645438a"}},{"code_sha256_prefix":"1a3baf406bc7de64","entry":"get_parameter_number","repo":"openbmb/omnilmm","repo_kind":"official","path":"finetune/finetune.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/finetune/finetune.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":"1a3baf406bc7de64"}},{"code_sha256_prefix":"1dd2ab205689d45b","entry":"identity_func","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/model/utils.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/model/utils.py","link_basis":"plan_row","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":"invariant","behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"1dd2ab205689d45b"}},{"code_sha256_prefix":"160038ef11a44314","entry":"parse_object_list","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/eval_gpt_obj_halbench.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/eval_gpt_obj_halbench.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":"160038ef11a44314"}},{"code_sha256_prefix":"763f6fe20bd52fbb","entry":"parse_score","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/eval_gpt_review_llava.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/eval_gpt_review_llava.py","link_basis":"harvester_set","language":"python","status":"ran_violates","verification_level":1,"contract_check":"VIOLATES","metamorphic_tier":"well_formed","behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"763f6fe20bd52fbb"}},{"code_sha256_prefix":"f7745bc5cafc18b6","entry":"preprocess_coh_results","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/eval_gpt_obj_halbench.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/eval_gpt_obj_halbench.py","link_basis":"plan_row","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":"f7745bc5cafc18b6"}},{"code_sha256_prefix":"f8c65c92ec213846","entry":"read_json","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/change_mmhal_predict_template.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/change_mmhal_predict_template.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":"f8c65c92ec213846"}},{"code_sha256_prefix":"2a1ffa5243150412","entry":"upload_img","repo":"openbmb/omnilmm","repo_kind":"official","path":"web_demos/web_demo.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/web_demos/web_demo.py","link_basis":"plan_row","language":"python","status":"ran_fixture","verification_level":1,"contract_check":"RAISES","metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"2a1ffa5243150412"}},{"code_sha256_prefix":"a32cd6f57d6772ae","entry":"conversation_to_ids","repo":"openbmb/omnilmm","repo_kind":"official","path":"finetune/dataset.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/finetune/dataset.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":"a32cd6f57d6772ae"}},{"code_sha256_prefix":"d6c1484ab50b4e26","entry":"conversation_to_ids_minicpm","repo":"openbmb/omnilmm","repo_kind":"official","path":"finetune/dataset.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/finetune/dataset.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":"d6c1484ab50b4e26"}},{"code_sha256_prefix":"4212780bf7da9ddd","entry":"data_collator","repo":"openbmb/omnilmm","repo_kind":"official","path":"finetune/dataset.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/finetune/dataset.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":"4212780bf7da9ddd"}},{"code_sha256_prefix":"2884deadbb637c3e","entry":"get_2d_sincos_pos_embed_from_grid","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/model/resampler.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/model/resampler.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":"2884deadbb637c3e"}},{"code_sha256_prefix":"77010a13f8bd8d11","entry":"get_eval","repo":"rlhf-v/rlhf-v","repo_kind":"listed","path":"eval/gpt4_grpc.py","file_url":"https://github.com/rlhf-v/rlhf-v/blob/HEAD/eval/gpt4_grpc.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":"77010a13f8bd8d11"}},{"code_sha256_prefix":"b474c783b31ec8ac","entry":"omni_preprocess","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/train/train_utils.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/train/train_utils.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":"b474c783b31ec8ac"}},{"code_sha256_prefix":"37899f22fb191b37","entry":"pretty_print_semaphore","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/utils.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/utils.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":"37899f22fb191b37"}},{"code_sha256_prefix":"f9939a84b9a65279","entry":"violates_moderation","repo":"openbmb/omnilmm","repo_kind":"official","path":"omnilmm/utils.py","file_url":"https://github.com/openbmb/omnilmm/blob/HEAD/omnilmm/utils.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":"f9939a84b9a65279"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}