VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon

  • 类型：arxiv
  • 标识：2607.01804
  • 链接：https://arxiv.org/abs/2607.01804
  • 主分类：multimodal
  • 形态：method
  • 被引：3
  • 被引来源：Semantic Scholar
  • S2被引：3
  • OpenAlex被引：0
  • 影响力被引：0
  • TLDR：VLA-Corrector, a lightweight corrective inference framework for action-chunked VLA policies, introduces a lightweight Latent-space Vision Monitor that continuously compares predicted and actual visual feature evolution, enabling online detection of visual dynamics deviations and mitigates the trade-off imposed by static horizons between execution robustness and policy-call frequency.
  • OpenAlex ID：W7167220322
  • OpenAlex DOI：10.48550/arxiv.2607.01804
  • DOI：10.48550/arxiv.2607.01804
  • DOI来源：OpenAlex
  • 开放获取：green
  • 开放获取链接：https://doi.org/10.48550/arxiv.2607.01804
  • OpenAlex更新：2026-07-19
  • 副分类：llm-infra
  • 待LLM分类：否
  • 标题中文：VLA-Corrector：面向自适应动作时域的轻量级检测-修正推理
  • TLDR中文：VLA-Corrector：面向动作分块 VLA 策略的轻量级修正推理框架；引入轻量的潜空间视觉监控器，持续比对预测与实际视觉特征演化，可在线检测视觉动态偏差，缓解静态时域在执行鲁棒性与策略调用频率之间的权衡。
  • 来源文件：
  • /inbox/tom/_candidates/2026-07-06-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]