Research map / Vision-language and multimodal / Multimodal LLMs
IQA-T1: Tool-based Visual Evidence Reasoning for Image Quality Assessment
Jinjian Wu, Jiaqi Tang, Wei Wei, Yingying Yan, Jianmin Chen, Botong Geng, Lei Zhang, Qifeng Chen
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 tool-generated visual evidence to SFT (ablation) | IQA-T1 SFT · Average of 7 IQA datasets | PLCC: 0.705 -> 0.756 | Improved |
| add tool-generated visual evidence to SFT (ablation) | IQA-T1 SFT · Average of 7 IQA datasets | SRCC: 0.680 -> 0.738 | Improved |
| add tool usage reward in RL (ablation) | IQA-T1 · Average of 7 IQA datasets | PLCC: 0.756 -> 0.790 | Improved |
| add repetition penalty in RL (ablation) | IQA-T1 · Average of 7 IQA datasets | PLCC: 0.756 -> 0.789 | Improved |
| add repetition penalty to model with tool usage reward (ablation) | IQA-T1 · Average of 7 IQA datasets | PLCC: 0.790 -> 0.795 | Improved |
| add tool usage reward to model with repetition penalty (ablation) | IQA-T1 · Average of 7 IQA datasets | SRCC: 0.775 -> 0.784 | Improved |
| dynamic tool invocation instead of all 15 tools (ablation) | IQA-T1 · KonIQ | PLCC: 0.865 -> 0.942 | Improved |
| dynamic tool invocation instead of fixed 3 tools (ablation) | IQA-T1 · KonIQ | PLCC: 0.895 -> 0.942 | Improved |
| dynamic tool invocation instead of random 3 tools (ablation) | IQA-T1 · KonIQ | PLCC: 0.871 -> 0.942 | Improved |
| add tool evidence to base Qwen3-VL-4B (ablation) | Qwen3-VL-4B · KonIQ | PLCC: 0.785 -> 0.942 | Improved |
| dynamic tool invocation instead of all tools (ablation) | IQA-T1 · KonIQ | avg inference time (s): 6.06 -> 5.09 | Improved |
| use Qwen3-VL-4B backbone instead of Qwen3-VL-2B (ablation) | IQA-T1 · KonIQ | PLCC: 0.929 -> 0.942 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Zoom-IQA — Zoom-IQA: Image Quality Assessment with Reliable Region-Aware Reasoning (ECCV 2026) · better on 3 of 3 shared results
- DeQA — DEQA: Descriptions Enhanced Question-Answering Framework for Multimodal Aspect-Based Sentiment Analysis (AAAI 2025) · better on 2 of 2 shared results
- Q-Insight — Q-Insight: Understanding Image Quality via Visual Reinforcement Learning (NeurIPS 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- IQA-Spider: Unifying Multi-Granularity Image Quality Assessment with Reasoning, Grounding and Referring (ICML 2026 · similarity 0.90)
- Adaptive Image Quality Assessment via Teaching Large Multimodal Model to Compare (NeurIPS 2024 · similarity 0.89)
- VisualQuality-R1: Reasoning-Induced Image Quality Assessment via Reinforcement Learning to Rank (NeurIPS 2025 · similarity 0.89)
- Reasoning as Representation: Rethinking Visual Reinforcement Learning in Image Quality Assessment (ICLR 2026 · similarity 0.89)
- Revisiting MLLM Based Image Quality Assessment: Errors and Remedy (AAAI 2026 · similarity 0.89)
- Grounding-IQA: Grounding Multimodal Language Model for Image Quality Assessment (ICLR 2026 · similarity 0.89)
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.
