AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

DyGRO-VLA: Cross-Task Scaling of Vision–Language–Action Models via Dynamic Grouped Residual Optimization

Sixu Lin, Yunpeng Qing, Litao Liu, Ming Zhou, Ruixing Jin, Xiaoyi Fan, Guiliang Liu

ICML 2026 · Reinforcement learning › Visual and representation RL · approach: Robust visual encoders for RL · 0 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
add reinforcement fine-tuning (online MoRR) to offline baseDyGRO-VLA · LIBERO Avgsuccess rate (%): 92.7 -> 97.1Improved
add reinforcement fine-tuning to offline baseDyGRO-VLA · LIBERO-Longsuccess rate (%): 85.2 -> 95.0Improved
increase mixture experts from 1 to 8 (ablation)DyGRO-VLA MoRR · LIBERO Avgsuccess rate (%): 93.4 -> 97.1Improved
increase experts from 4 to 8 (ablation)DyGRO-VLA MoRR · LIBERO Avgsuccess rate (%): 97.0 -> 97.1No clear effect
add information bottleneck objective (ablation)DyGRO-VLA · LIBERO Avgsuccess rate (%): 96.8 -> 97.1Improved
add InfoNCE contrastive objective (ablation)DyGRO-VLA · LIBERO Avgsuccess rate (%): 94.3 -> 97.1Improved
add InfoNCE contrastive objective (ablation)DyGRO-VLA · LIBERO-Longsuccess rate (%): 90.4 -> 95.0Improved
add difficulty-aware sampling (ablation)DyGRO-VLA · LIBERO Avgsuccess rate (%): 96.4 -> 97.1Improved
replace shared MLP residual head with MoRR dynamic grouping (ablation)DyGRO-VLA · Gradient cosine similarity (Open/Close tasks)gradient similarityImproved
update VLM backbone with LoRA in online RL (ablation)VLM (LoRA) + Action Expert · LIBERO online trainingsuccess rateHurt
use standard residual RL instead of MoRR (ablation)Standard Residual RL · LIBERO online trainingsuccess rateHurt
DyGRO-VLA vs OpenVLA-oft RFT on Sim2RealDyGRO-VLA · RoboTwin2 Avg (Real)success rate (%): 55.0 -> 57.5Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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