Research map / Vision-language and multimodal / Cross-modal learning
UniGame: Turning a Unified Multimodal Model Into Its Own Adversary
Zhaolong Su, Wang Lu, Hao Chen, Sharon Li, Jindong Wang
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 UniGame self-adversarial post-training | Janus-Pro-7B · Consistency (UnifiedBench + WISE) | consistency score: 63.66 -> 68.32 | Improved |
| use UniGame instead of plain SFT | Janus-Pro-7B · Consistency (UnifiedBench + WISE) | consistency score: 64.72 -> 68.32 | Improved |
| add UniGame self-adversarial post-training | Janus-Pro-7B · Understanding overall | average score: 71.4 -> 75.0 | Improved |
| use UniGame instead of plain SFT | Janus-Pro-7B · Understanding overall | average score: 71.9 -> 75.0 | Improved |
| add plain SFT post-training | Janus-Pro-7B · Understanding overall | average score: 71.4 -> 71.9 | No clear effect |
| add UniGame self-adversarial post-training | Janus-Pro-7B · VQAv2 (test) | accuracy: 78.2 -> 83.4 | Improved |
| add UniGame self-adversarial post-training | Janus-Pro-7B · POPE | accuracy: 87.4 -> 89.6 | Improved |
| add UniGame self-adversarial post-training | Janus-Pro-7B · GenEval | overall score: 0.80 -> 0.82 | Improved |
| add plain SFT post-training | Janus-Pro-7B · GenEval | overall score: 0.80 -> 0.81 | No clear effect |
| UniGame (trained mainly on understanding) versus RecA with explicit generation post-training | Janus-Pro-7B · GenEval | overall score: 0.82 vs. 0.86 | Hurt |
| add UniGame self-adversarial post-training | Janus-Pro-7B · NaturalBench (OOD robustness) | group accuracy: +4.8% | Improved |
| add UniGame self-adversarial post-training | Janus-Pro-7B · AdVQA (adversarial robustness) | accuracy: +6.2% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- OmniGen2 — OmniGen2: Towards Instruction-Aligned Multimodal Generation (CVPR 2026) · better on 2 of 2 shared results
- Show-o — Show-o: One Single Transformer to Unify Multimodal Understanding and Generation (ICLR 2025) · better on 1 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
- UniM: A Unified Any-to-Any Interleaved Multimodal Benchmark (CVPR 2026 · similarity 0.87)
- Show-o: One Single Transformer to Unify Multimodal Understanding and Generation (ICLR 2025 · similarity 0.86)
- Learning to Generate via Understanding: Understanding-Driven Intrinsic Rewarding for Unified Multimodal Models (CVPR 2026 · similarity 0.86)
- RealUnify: Do Unified Models Truly Benefit from Unification? A Comprehensive Benchmark (CVPR 2026 · similarity 0.86)
- Generation Enhances Understanding in Unified Multimodal Models via Multi-Representation Generation (ICML 2026 · similarity 0.86)
- DuoGen: Towards Autonomous Interleaved Multimodal Generation (CVPR 2026 · similarity 0.86)
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.
