{"paper":{"title":"Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"LMMs achieve better visual scoring by predicting discrete text-defined rating levels instead of numerical scores.","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Annan Wang, Chaofeng Chen, Chunyi Li, Erli Zhang, Guangtao Zhai, Haoning Wu, Liang Liao, Qiong Yan, Weisi Lin, Weixia Zhang, Wenxiu Sun, Xiongkuo Min, Yixuan Gao, Zicheng Zhang","submitted_at":"2023-12-28T16:10:25Z","abstract_excerpt":"The explosion of visual content available online underscores the requirement for an accurate machine assessor to robustly evaluate scores across diverse types of visual contents. While recent studies have demonstrated the exceptional potentials of large multi-modality models (LMMs) on a wide range of related fields, in this work, we explore how to teach them for visual rating aligned with human opinions. Observing that human raters only learn and judge discrete text-defined levels in subjective studies, we propose to emulate this subjective process and teach LMMs with text-defined rating level"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The proposed Q-Align achieves state-of-the-art performance on image quality assessment (IQA), image aesthetic assessment (IAA), as well as video quality assessment (VQA) tasks under the original LMM structure. With the syllabus, we further unify the three tasks into one model, termed the OneAlign.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That training LMMs with discrete text-defined levels emulates human subjective judgment processes more effectively than direct numerical score regression, leading to better performance without architectural changes or extra data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Q-Align trains LMMs on discrete text-defined levels for visual scoring, achieving SOTA on IQA, IAA, and VQA while unifying the tasks in OneAlign.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LMMs achieve better visual scoring by predicting discrete text-defined rating levels instead of numerical scores.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"1a1861089861b0b862db8716832fe34666ffb4075a526d038849951b686c6111"},"source":{"id":"2312.17090","kind":"arxiv","version":1},"verdict":{"id":"ef03bab8-e2e6-48e8-be61-5f777a6671b1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T16:31:49.640259Z","strongest_claim":"The proposed Q-Align achieves state-of-the-art performance on image quality assessment (IQA), image aesthetic assessment (IAA), as well as video quality assessment (VQA) tasks under the original LMM structure. With the syllabus, we further unify the three tasks into one model, termed the OneAlign.","one_line_summary":"Q-Align trains LMMs on discrete text-defined levels for visual scoring, achieving SOTA on IQA, IAA, and VQA while unifying the tasks in OneAlign.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That training LMMs with discrete text-defined levels emulates human subjective judgment processes more effectively than direct numerical score regression, leading to better performance without architectural changes or extra data.","pith_extraction_headline":"LMMs achieve better visual scoring by predicting discrete text-defined rating levels instead of numerical scores."},"references":{"count":293,"sample":[{"doi":"","year":null,"title":"FirstName LastName , title =","work_id":"d9cab501-317f-4237-9e32-b5ead5964402","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"FirstName Alpher , title =","work_id":"42297990-8783-41a1-b0fa-8ccdbf630852","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Journal of Foo , volume = 13, number = 1, pages =","work_id":"65a8b3d0-af84-4f68-87eb-101c85ab18b2","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Journal of Foo , volume = 14, number = 1, pages =","work_id":"b3089947-bd36-4a24-9199-cc535e299537","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"FirstName Alpher and FirstName Gamow , title =","work_id":"caed320b-7cdc-41ca-bb08-00fb14feec62","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":293,"snapshot_sha256":"681418912eaa42e4611d22f6581da11de675b162729c34822c93850dc341d530","internal_anchors":9},"formal_canon":{"evidence_count":3,"snapshot_sha256":"7b9b449cf34fb4622e264d8a1a0b046365f58616194452f8a731628681551281"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}