ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models

类型：arxiv
标识：2609.18487
链接：https://arxiv.org/abs/2609.18487
主分类：multimodal
形态：position
TLDR：Action tokenizers play a central role in autoregressive vision-language-action (VLA) models, determining both the targets for policy training and the executable commands recovered from predicted tokens. Their fidelity is commonly evaluated using pointwise reconstruction metrics such as mean squared error (MSE), yet small individual errors do not fully characterize how faithfully action adjustments across demonstrations are preserved. After compression, similar actions may still cluster around a representative motion, while the adjustments needed for different contexts are diminished, distorted
待LLM分类：否
来源文件：
/inbox/tom/_candidates/2026-09-17-agent-rag-longcontext-candidates.json

2026-09-12

笔记精读与批判 v2 · MMProLong — 长上下文 VLM 继续预训练（LongPT）配方 · 2026-09-14 (重写覆盖)

2026-09-14

笔记 MMProLong:长上下文视觉语言模型的有效续训练(精读 · flyP)

2026-06-14

笔记精读：Scaling Beyond Context — 多模态 RAG for Document Understanding Survey

2026-09-14

仓库 leon2k2k2k/llm-feedback-alignment

2026-08-11

仓库 BradyFU/Awesome-Multimodal-Large-Language-Models

2026-08-14

arXiv

2609.18487

TLDR

Action tokenizers play a central role in autoregressive vision-language-action (VLA) models, determining both the targets for policy training and the executable commands recovered from predicted tokens. Their fidelity is commonly evaluated using pointwise reconstruction metrics such as mean squared error (MSE), yet small individual errors do not fully characterize how faithfully action adjustments across demonstrations are preserved. After compression, similar actions may still cluster around a representative motion, while the adjustments needed for different contexts are diminished, distorted

arXiv ↗ 中文全文

ActionPiece: Rethinking Action Tokenization for Autoregressive Vision-Language-Action Models

相关 · 按语义匹配