Research map / Vision-language and multimodal / Multimodal LLMs
UEval: A Benchmark for Unified Multimodal Generation
Bo Li, Yida Yin, Wenhao Chai, Xingyu Fu, Zhuang Liu
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 |
|---|---|---|---|
| append GPT-5-Thinking reasoning trace to the prompt (ablation) | GPT-5-Instant · UEval image generation | image quality | Improved |
| append GPT-5-Thinking reasoning trace to the prompt (ablation) | Gemini-2.5-Flash · UEval image generation | image quality | Improved |
| append GPT-5-Thinking reasoning trace to the prompt (ablation) | BAGEL · UEval | generation quality | No clear effect |
| use reasoning model GPT-5-Thinking instead of GPT-5-Instant | GPT-5 · UEval avg | rubric score: 65.2 -> 66.4 | Improved |
| use reasoning model GPT-5-Thinking instead of GPT-5-Instant | GPT-5 · UEval Paper | rubric score: 55.1 -> 51.9 | Hurt |
| use proprietary GPT-5-Thinking instead of best open-source Emu3.5 | unified models · UEval avg | rubric score: 49.1 -> 66.4 | Improved |
| generate text in unified models instead of images | GPT-5-Thinking · UEval | rubric score: 49.1 -> 83.8 | Improved |
| open-source unified model text versus text-only model | Emu3.5 · UEval text | text score: 64.6 -> 81.2 | Hurt |
| multi-step planning tasks versus knowledge tasks | GPT-5-Thinking · UEval | rubric score: 78.0 -> 67.8 | Hurt |
| use Seed1.6-Vision as judge instead of Gemini-2.5-Pro (ablation) | GPT-5-Thinking responses · UEval Art | rubric score: 67.8 -> 75.2 | Mixed |
| use GLM-4.1V-Thinking as judge instead of Gemini-2.5-Pro (ablation) | GPT-5-Thinking responses · UEval Tech | rubric score: 57.0 -> 74.3 | Mixed |
| use Qwen3-VL-235B-Thinking as judge instead of Gemini-2.5-Pro (ablation) | GPT-5-Thinking responses · UEval Space | rubric score: 84.0 -> 81.8 | No clear effect |
Closest papers by meaning
- LENS: Multi-level Evaluation of Multimodal Reasoning with Large Language Models (ICLR 2026 · similarity 0.91)
- ROVER: Benchmarking Reciprocal Cross-Modal Reasoning for Omnimodal Generation (ICLR 2026 · similarity 0.91)
- RPTS: Tree-Structured Reasoning Process Scoring for Faithful Multimodal Evaluation (AAAI 2026 · similarity 0.90)
- MME-Unify: A Comprehensive Benchmark for Unified Multimodal Understanding and Generation Models (ICLR 2026 · similarity 0.90)
- UniVLR: Unifying Text and Vision in Visual Latent Reasoning for Multimodal LLMs (NeurIPS 2026 · similarity 0.89)
- UPME: An Unsupervised Peer Review Framework for Multimodal Large Language Model Evaluation (CVPR 2025 · 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.
