AI Paper Atlas

Research map / Vision-language and multimodal / Robot action and navigation

Trust Your Instincts: Confidence-Driven Test-Time RL for Vision-Language-Action Models

Chen Siyao, Jiakang Yuan, Jiaxin Wang, Tao Chen

ECCV 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models · 2 citations (OpenAlex)

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 changedWhereMeasuredEffect
apply T2VLA self-reward test-time RLOpenVLA-OFT · LIBERO averagesuccess rate: +6.2Improved
apply T2VLA self-reward test-time RLpi0 · LIBERO averagesuccess rate: +24.2Improved
apply T2VLA self-reward test-time RLpi0.5 · LIBERO averagesuccess rate: +8.0Improved
apply T2VLA self-reward test-time RLOpenVLA-OFT · RoboTwin 2.0 averagesuccess rate: +21.3Improved
add the global expert pool to local expert only (ablation)T2VLA · LIBERO-Longsuccess rate: 94.5 to 95.3Improved
add the local pseudo-expert to global pool only (ablation)T2VLA · LIBERO-Longsuccess rate: 93.0 to 95.3Improved
set pool capacity K=5 instead of 3 (ablation)T2VLA · LIBERO-Longsuccess rate: 93.4 to 95.3Improved
set pool capacity K=5 instead of 10 (ablation)T2VLA · LIBERO-Longsuccess rate: 91.4 to 95.3Improved
use min-max confidence normalization instead of static weighting (ablation)T2VLA · LIBERO-Longsuccess rate: 88.3 to 95.3Improved
use min-max confidence normalization instead of max routing (ablation)T2VLA · LIBERO-Longsuccess rate: 87.5 to 95.3Improved
use min-max normalization instead of sigmoid gate (ablation)T2VLA · LIBERO-Longsuccess rate: 92.6 to 95.3Improved
apply T2VLA to a 1-shot SFT policyOpenVLA-OFT · LIBERO-Long (1-shot)success rate: 17.3 to 11.0Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy