AI Paper Atlas

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

A Generalist Pair-wise Progress Critic Model for Vision-Language-Action Robots

Qi Zhang, Shaopeng Zhai, Shengzhe Zhang, Litao Liu, Tianyi Zhang, Fuxian Huang, Ming Zhou

ICML 2026 · Vision-language and multimodal › Robot action and navigation · approach: Vision-language-action foundation models

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
add pair-wise progress understanding pretraining (ablation)VLAC action generation · Real-world manipulation (5 tasks)average success rate (%): 16 -> 75Improved
use semantic-space numeric actions instead of FAST tokenizer (ablation)VLAC · Real-world manipulation (5 tasks)average success rate (%): 66 -> 75Improved
use VLAC instead of pi0 under lighting disturbanceVLA policy · Real-world manipulation, lighting transferaverage success rate (%): 3 -> 57Improved
VLAC instead of pi0 in-distributionVLA policy · Real-world manipulation (5 tasks)average success rate (%): 27 -> 75Improved
pair-wise progress prediction instead of single-point estimation (ablation)VLAC-8b critic · eight critic datasetsVOCImproved
add one reference example (in-context one-shot) (ablation)VLAC-8b critic · RoboNetVOC-F1: 0 -> improves significantlyImproved
add Ego4D human video to training (ablation)VLAC-8b critic · EgoDex and RoboFACVOC-F1Improved
add large-scale VLAC pretraining before post-training (ablation)critic post-training · EgoDexVOC-F1: 0.24 -> 0.68Improved
add large-scale VLAC pretraining before post-training (ablation)critic post-training · RoboFAC-failVOC-F1: 0.21 -> 0.37Improved
mix 10% failure trajectories into RoboFAC-success training (ablation)VLAC · RoboFAC-successVOC-F1: 0.99 -> 0.95Hurt
use dense progress-based reward instead of sparse reward in real-world RLVLAC with PPO · Real-world RLsuccess rate: 30% -> 90%Improved

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