PulseAugur
中
实时 07:32:43
English(EN) How Many Demonstrations Does a VLA Need? On the Sample Efficiency of VLA Finetuning

研究发现VLA模型在微调时表现出不同的样本效率

一项新的研究论文调查了视觉语言动作(VLA)模型在演示数据上进行微调时的样本效率。该研究揭示了包括 $\pi_0$、$\pi_{0.5}$ 和 GR00T N1-N1.7 在内的六个 VLA 模型在四个 LIBERO 基准套件上的适应效率存在显著差异。值得注意的是,$\pi_{0.5}$ 表现出优越的样本效率,比 $\pi_0$ 需要更少的演示,并且在有限数据下实现了更高的成功率。研究还发现,在 GR00T 模型中,尽管 N1.6 和 N1.7 的零样本性能更强,但 N1.5 的效率更高。 AI

影响 强调模型架构和预训练选择对微调效率有显著影响,为未来的VLA开发提供指导。

排序理由 研究论文,详细介绍了VLA模型样本效率的发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

研究发现VLA模型在微调时表现出不同的样本效率

本文如何被排名

Signal score
30 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
研究论文,详细介绍了VLA模型样本效率的发现。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
paper, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

  1. arXiv cs.CV TIER_1 English(EN) · Rahul Chand, Miguel Villa Floran ·

    VLA需要多少次演示?关于VLA微调的样本效率

    arXiv:2611.09964v2 Announce Type: replace Abstract: Pretrained vision language action (VLA) models show strong zero-shot performance but they are rarely deployed directly and are often first post-trained on demonstrations from the target setting. Yet VLAs are typically compared b…