AI Paper Atlas

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

Robust Fine-tuning of Vision-Language-Action Robot Policies via Parameter Merging

Yajat Yadav, Zhiyuan Zhou, Andrew Wagenmaker, Karl Pertsch, Sergey Levine

ICLR 2026 · Vision-language and multimodal › Robot action and navigation · approach: Multimodal prompt-driven manipulation

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
Interpolate finetuned and pretrained weights (RETAIN)finetuned pi0-FAST-DROID · DROID whiteboard and plates, OODsuccess rate: baselines 30-50% -> close to 80% (whiteboard), >60% (plates)Improved
Merge weights with pretrained policyfinetuned policy · DROID and LIBERO, ID target tasksuccess rateNo clear effect
Merge weights with pretrained policyfinetuned policy · Generalist tasks (44 DROID, 20 LIBERO)success rateImproved
Add co-finetuning before merging (RETAIN-co-FT vs RETAIN-task-FT) (ablation)RETAIN · DROID and LIBERO, generalist and OODsuccess rateImproved
Use more pretraining data for the base policy (ablation)RETAIN-co-FT · DROID plates, OODsuccess gain over co-FTImproved
Use more pretraining data for the base policy (ablation)RETAIN-co-FT · DROID plates, IDsuccess rateNo clear effect
Merge only language-model parameters instead of all parameters (ablation)RETAIN · LIBERO, OODsuccess rateNo clear effect
Set alpha_l to 1 (no merging of language model) instead of 0.8 (ablation)RETAIN modality-specific · LIBERO mugs-on-plates, OODsuccess rateHurt
Set merging coefficient alpha too close to pretrained model (ablation)RETAIN · LIBERO, OODsuccess rate: near 0 performanceHurt
Lower the learning rate or tune gradient steps in task-FT (ablation)Task-FT · LIBERO, OODsuccess rateHurt
Apply RETAIN in sequential skill acquisition instead of co-FTDROID plates then whiteboard · ID and OODsuccess rateImproved
Use RETAIN with a less general base model (LIBERO)RETAIN · LIBERO OOD vs DROID OODOOD gain over Co-FTMixed

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