Research map / Vision-language and multimodal / Robot action and navigation
FASTer: Toward Powerful and Efficient Autoregressive Vision–Language–Action Models with Learnable Action Tokenizer and Block-wise Decoding
Yicheng Liu, Shiduo Zhang, Zibin Dong, Baijun Ye, Tianyuan Yuan, Xiaopeng Yu, Linqi Yin, Chenhao Lu, Junhao Shi, Luca Jiang-Tao Yu, Liangtao Zheng, Jingjing Gong and 3 more
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 block-wise autoregressive decoding (ablation) | FASTer · LIBERO (average) | success rate (%): 95.4 -> 97.9 | Improved |
| add block-wise autoregressive decoding (ablation) | FASTer · Simpler-Bridge (average) | success rate (%): 81.0 -> 87.9 | Improved |
| replace token-wise AR with block-wise decoding (ablation) | FASTer · LIBERO | latency (ms): 323 -> 140 | Improved |
| replace token-wise AR with block-wise decoding (ablation) | FASTer · LIBERO | success rate (%): 95.5 -> 96.7 | Improved |
| add pretrained action expert (ablation) | FASTer · LIBERO | success rate (%): 95.5 -> 97.9 | Improved |
| add action expert without pretraining (ablation) | FASTer · LIBERO | success rate (%): 95.5 -> 94.8 | Hurt |
| add action expert trained from scratch (ablation) | FASTer · simpler-widow | success rate (%): 75.6 -> 23.6 | Hurt |
| add pretrained action expert (ablation) | FASTer · simpler-widow | success rate (%): 75.6 -> 87.9 | Improved |
| use TAAE tokenizer instead of CNN (ablation) | FASTerVQ · LIBERO | success rate (%): 96.2 -> 97.9 | Improved |
| increase codebook size from 4096 to 8192 (ablation) | FASTerVQ · LIBERO | success rate (%): 97.9 -> 96.3 | Hurt |
| increase residual codebooks from 1 to 3 (ablation) | FASTerVQ · LIBERO | success rate (%): 93.4 -> 97.9 | Improved |
| increase residual codebooks from 3 to 8 (ablation) | FASTerVQ · LIBERO | success rate (%): 97.9 -> 96.6 | Hurt |
Closest papers by meaning
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.93)
- ActionUNet: Improving Robustness of VLA Models with Efficient Multi-scale Fine-tuning (NeurIPS 2026 · similarity 0.93)
- VLA-Cache: Efficient Vision-Language-Action Manipulation via Adaptive Token Caching (NeurIPS 2025 · similarity 0.92)
- NS-VLA: Towards Neuro-Symbolic Vision-Language-Action Models (NeurIPS 2026 · similarity 0.92)
- VLA-REPLICA: A Low-Cost, Reproducible Benchmark for Real-World Evaluation of Vision-Language-Action Models (NeurIPS 2026 · similarity 0.91)
- EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models (NeurIPS 2025 · similarity 0.91)
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.
