Research map / Vision-language and multimodal / Multimodal LLMs
ILLUME: Illuminating Your LLMs to See, Draw, and Self-Enhance
Chunwei Wang, Guansong Lu, Junwei Yang, Runhui Huang, Jianhua Han, Lu Hou, Wei Zhang, Hang Xu
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 50K self-assessment data in stage 3 (ablation) | ILLUME · MME-P | score: 1358.6 -> 1446.7 | Improved |
| add self-assessment data (ablation) | ILLUME · MMBench | accuracy: 61.7 -> 63.1 | Improved |
| add self-assessment data (ablation) | ILLUME · MM-Vet | accuracy: 27.4 -> 29.0 | Improved |
| add self-assessment data (ablation) | ILLUME · SEED | accuracy: 65.0 -> 66.0 | Improved |
| add self-assessment data (ablation) | ILLUME · POPE | accuracy: 86.4 -> 86.1 | No clear effect |
| add self-assessment data (ablation) | ILLUME · GenEval overall | accuracy: 0.56 -> 0.59 | Improved |
| add self-assessment data (ablation) | ILLUME · GenEval counting | accuracy: 0.35 -> 0.43 | Improved |
| add self-assessment data (ablation) | ILLUME · GenEval position | accuracy: 0.34 -> 0.33 | No clear effect |
| use semantic tokenizer instead of reconstruction tokenizer (ablation) | ILLUME MLLM · text-to-image training (20M) | training loss / generation | Improved |
| joint understanding and generation training vs specialist (ablation) | ILLUME · MME-P | score: 1339.0 -> 1358.6 | No clear effect |
| joint training vs generation-only (ablation) | ILLUME · GenEval | accuracy: 0.58 -> 0.56 | No clear effect |
| joint training vs understanding-only (ablation) | ILLUME · MMBench | accuracy: 60.9 -> 61.6 | No clear effect |
Later papers that compared with this method
4 later accepted papers report ILLUME+ 3B as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Show Me Examples: Inferring Visual Concepts from Image Sets (ECCV 2026 · better on 3 of 3 shared results)
- UniVR: Thinking in Visual Space for Unified Visual Reasoning (NeurIPS 2026 · better on 1 of 1 shared result)
- FOCUS: Unified Vision-Language Modeling for Interactive Editing Driven by Referential Segmentation (NeurIPS 2025 · better on 17 of 19 shared results)
- FUDOKI: Discrete Flow-based Unified Understanding and Generation via Kinetic-Optimal Velocities (NeurIPS 2025 · better on 0 of 2 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- Janus — Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation (CVPR 2025) · better on 4 of 5 shared results
- VILA-U (384) — VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation (ICLR 2025) · better on 0 of 1 shared result
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
- SILMM: Self-Improving Large Multimodal Models for Compositional Text-to-Image Generation (CVPR 2025 · similarity 0.90)
- Implicit Multimodal Alignment: On the Generalization of Frozen LLMs to Multimodal Inputs (NeurIPS 2024 · similarity 0.89)
- TokenSwap: Benchmarking and Reducing the Modality Gap in Multimodal LLMs (NeurIPS 2026 · similarity 0.89)
- Analyzing Fine-Grained Alignment and Enhancing Vision Understanding in Multimodal Language Models (NeurIPS 2025 · similarity 0.89)
- Beyond Next-Token Alignment: Distilling Multimodal Large Language Models via Token Interactions (ICML 2026 · similarity 0.89)
- Vision as a Dialect: Unifying Visual Understanding and Generation via Text-Aligned Representations (NeurIPS 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.
