AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Reference Guided Skill Discovery

Seungeun Rho, Aaron Trinh, Danfei Xu, Sehoon Ha

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: Unsupervised skill discovery

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
ground latent space with contrastive pretraining on reference motions (RGSD vs plain DIAYN) (ablation)DIAYN · imitation on run motionsCartesian error (cm): 52.8 -> 7.7Improved
RGSD frame-level latent reward vs GAIL-based imitation (Meta-Motivo) on fidelityMeta-Motivo · imitation, 5 motionsCartesian error: outperforms in 4 of 5 tasks, e.g. Backward 8.6 -> 6.7Improved
RGSD vs Meta-Motivo on naturalnessRGSD · imitation, 5 motionsMotion FID: Run 4.9 vs 9.4 (RGSD)Hurt
train imitation and discovery in parallel with neighborhood latent samplingRGSD · skill discovery near reference motionsFID stability: CALM walking 1.4 -> 15.5 vs stable RGSDImproved
increase vMF concentration kappa at test time (ablation)RGSD · three reference motions, kappa 20/100/1000behavioral diversityMixed
RGSD high-level policy on goal reachingRGSD · GoalReaching freestyleconvergence speedImproved
RGSD on style-conditioned goal reaching (sidestep, backward)RGSD · GoalReaching stylesstyle adherence FIDImproved
extend RGSD with METRA backbone instead of DIAYN (ablation)RGSD with METRA · repetitive motionsrewardHurt

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