Research map / Vision-language and multimodal / Robot action and navigation
CombatVLA: An Efficient Vision-Language-Action Model for Combat Tasks in 3D Action Role-Playing Games
Peng Chen, Pi Bu, Yingyao Wang, Xinyi Wang, Ziming Wang, Jie Guo, Yingxiu Zhao, Qi Zhu, Jun Song, Siran Yang, Jiamang Wang, Bo Zheng
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 frame-level Frames-AoT tuning (stage 2) after video-AoT (ablation) | CombatVLA · CUBench | average accuracy: 57.27 -> 61.43 | Improved |
| add truncated AoT tuning (stage 3) after stage 2 (ablation) | CombatVLA · CUBench | average accuracy: 61.43 -> 63.61 | Improved |
| add truncated AoT (stage 3) (ablation) | CombatVLA · CUBench reasoning | accuracy: 62.86 -> 69.71 | Improved |
| add truncated AoT (stage 3) (ablation) | CombatVLA · CUBench | time per call (s): 3.73 -> 1.85 | Improved |
| add truncated AoT (stage 3) (ablation) | CombatVLA · CUBench comprehension | accuracy: 62.25 -> 60.29 | Hurt |
| add modality contrastive loss L_con (ablation) | CombatVLA · CUBench | average accuracy: 61.58 -> 63.61 | Improved |
| add action alignment loss L_align (ablation) | CombatVLA · CUBench | average accuracy: 61.64 -> 63.61 | Improved |
| add L_con (ablation) | CombatVLA · CUBench reasoning | accuracy: 63.14 -> 69.71 | Improved |
| add L_align (ablation) | CombatVLA · CUBench reasoning | accuracy: 63.71 -> 69.71 | Improved |
| add L_con (ablation) | CombatVLA · CUBench gathering | accuracy: 62.78 -> 60.83 | Hurt |
| truncate AoT output at TRUNC token | CombatVLA · Average AoT token length | tokens: 116.57 -> 43.10 | Improved |
| fine-tune Qwen2.5-VL-3B with AoT (CombatVLA) | Qwen2.5-VL-3B · CUBench | average accuracy: 55.87 -> 63.61 | Improved |
Closest papers by meaning
- VLA-OS: Structuring and Dissecting Planning Representations and Paradigms in Vision-Language-Action Models (NeurIPS 2025 · similarity 0.89)
- VLA-ATTC: Adaptive Test-Time Compute for VLA Models with Relative Action Critic Model (ICML 2026 · similarity 0.89)
- 3D-VLA: A 3D Vision-Language-Action Generative World Model (ICML 2024 · similarity 0.88)
- GameVerse: Can Vision-Language Models Learn from Video-based Reflection? (ICML 2026 · similarity 0.88)
- VS-Bench: Evaluating VLMs for Strategic Abilities in Multi-Agent Environments (CVPR 2026 · similarity 0.87)
- Vision-Zero: Scalable VLM Self-Improvement via Strategic Gamified Self-Play (ICLR 2026 · similarity 0.87)
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.
