Research map / Vision-language and multimodal / Multimodal LLMs
Depicting Beyond Scores: Advancing Image Quality Assessment through Multi-modal Language Models
Zhiyuan You, Zheyuan Li, Jinjin Gu, Zhenfei Yin, Tianfan Xue, Chao Dong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Task-2 brief templated comparison data to training (ablation) | DepictQA · BAPPS Traditional | comparison accuracy: 66.4 -> 79.7 | Improved |
| add Task-3 reasoning data to training (ablation) | DepictQA · M-BAPPS reasoning val | reasonable rate: N/A -> 41.0 (Task2 vs Task2+Task3) | Improved |
| add Task-1 description data without content description data (ablation) | DepictQA · BAPPS Traditional | comparison accuracy: 79.7 -> 78.1 | Hurt |
| add Task-1 description data (ablation) | DepictQA · M-BAPPS reasoning val | reasonable rate: 41.0 -> 45.0 | Improved |
| add IQA-unrelated content description data as regularization (ablation) | DepictQA · BAPPS Traditional | comparison accuracy: 78.1 -> 80.0 | Improved |
| add content description data (ablation) | DepictQA · M-BAPPS reasoning val | reasonable rate: 45.0 -> 53.0 | Improved |
| use unique tag per image instead of unified tag (ablation) | DepictQA · 50 reviewed reasoning responses | confusion rate: 24% -> 12% | Improved |
| use unique tag instead of image embedding or unique projector (ablation) | DepictQA · M-BAPPS reasoning val | reasonable rate: 26.0 / 40.0 -> 53.0 | Improved |
| use local feature map instead of single global CLIP feature (ablation) | DepictQA · M-BAPPS reasoning val | reasonable rate: 34.0 -> 53.0 | Improved |
| initialize from MLLM pre-trained on high-level perception tasks (ablation) | DepictQA (Vicuna-v1.5) · BAPPS Traditional | comparison accuracy: 80.0 -> 81.2 | Improved |
| vary LoRA rank from 12 to 32 (ablation) | DepictQA · BAPPS average | accuracy: 81.8 / 81.9 / 81.8 / 82.1 | No clear effect |
| use LoRA rank 8 (ablation) | DepictQA · BAPPS | accuracy: fail | Hurt |
Closest papers by meaning
- Teaching Large Language Models to Regress Accurate Image Quality Scores Using Score Distribution (CVPR 2025 · similarity 0.91)
- Revisiting MLLM Based Image Quality Assessment: Errors and Remedy (AAAI 2026 · similarity 0.90)
- Grounding-IQA: Grounding Multimodal Language Model for Image Quality Assessment (ICLR 2026 · similarity 0.90)
- IQA-Spider: Unifying Multi-Granularity Image Quality Assessment with Reasoning, Grounding and Referring (ICML 2026 · similarity 0.89)
- Q-Bench: A Benchmark for General-Purpose Foundation Models on Low-level Vision (ICLR 2024 · similarity 0.88)
- Adaptive Image Quality Assessment via Teaching Large Multimodal Model to Compare (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
