AI Paper Atlas

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

SkillNet: Hierarchical Skill Modeling for Compositional Generalization in Vision-Language Action Models

Senwei Xie, Yuntian Zhang, Zhenzhou Tan, Ruiping Wang, Pengwei Wang, Shanghang Zhang, Xilin CHEN

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 skill-centric MoE (SkillNet) to pi0.5pi0.5 · LIBERO-Skill zero-shotsuccess rate: 31.8 -> 47.8Improved
replace vanilla MoE with SCMoE (ablation)pi0.5 backbone · LIBERO-Skill zero-shotsuccess rate: 33.1 -> 47.8Improved
use vanilla MoE instead of nothing on pi0.5pi0.5 · LIBERO-Skill zero-shotsuccess rate: 31.8 -> 33.1No clear effect
add SkillNet to pi0.5 on real-world zero-shotpi0.5 · Real-world compositional taskssuccess rate: 42.5 -> 62.5Improved
add SkillNet few-shot transferpi0.5 · Real-world unseen behaviors 15-shotsuccess rate: +23.9% averageImproved
add SkillNet few-shot transferpi0.5 · RoboTwin-2.0 15-shotsuccess rate: +5.9%Improved
add SkillNet in-domainpi0.5 · LIBERO in-domainsuccess rate: 98.0 -> 99.1Improved
add SkillNet in-domain real-world long-horizonVanilla MoE · Real-world long-horizonsuccess rate: 50.0 -> 55.7Improved
add SkillNet on RoboCasapi0.5 and vanilla MoE · RoboCasasuccess rate: +1.7%Improved
add SCMoE instead of vanilla MoE (ablation)SkillNet · Zero-shot generalization ablationsuccess rateImproved
add hierarchical skill modeling (HSM) instead of flat skill IDs (ablation)SkillNet · Few-shot transfer ablationsuccess rateImproved
add router balance loss (ablation)SkillNet · Few-shot transfer ablationsuccess rate: 6.2% degradation when removedImproved

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