AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Saliency-Guided Representation with Consistency Policy Learning for Visual Unsupervised Reinforcement Learning

Jingbo Sun, Qichao Zhang, Songjun Tu, Xing Fang, Yupeng Zheng, Haoran Li, Ke Chen, Dongbin Zhao

CVPR 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 saliency-guided dynamics representation learning (decouple encoder from SR objective) (ablation)HILP + consistency policy · ExORL RND, Walkerreturn: 345 -> 439Improved
add saliency-guided dynamics representation learning (ablation)HILP + consistency policy · ExORL RND, Quadrupedreturn: 352 -> 485Improved
add consistency policy (ablation)SRCP with saliency representation · ExORL RND, Walkerreturn: 396 -> 439Improved
add consistency policy (ablation)SRCP with saliency representation · ExORL RND, Quadrupedreturn: 406 -> 485Improved
add saliency representation and consistency policy (ablation)HILP · ExORL RND, Cheetahreturn: 599 -> 602No clear effect
add SRCP components (ablation)HILP · ExORL RND, Jacoreturn: 34 -> 50Improved
use classifier-free guidance (omega=3 vs 0) (ablation)SRCP · ExORL RND, Quadrupedreturn: 234 -> 485Improved
increase guidance weight omega from 3 to 6 (ablation)SRCP · ExORL RND, Quadrupedreturn: 485 -> 455Hurt
add saliency loss weight beta=0.5 (vs 0) (ablation)SRCP · ExORL RND, Walkerreturn: 341 -> 439Improved
increase beta from 0.5 to 1 (ablation)SRCP · ExORL RND, Walkerreturn: 439 -> 396Hurt
plug SRCP into forward-backward (FB) learnerFB · ExORL RND, Walker and Quadruped (8 tasks)return: +23% to +490% across tasksImproved
replace HILP encoder objective with SRCP vs TACO/Premier-TACO contrastive representations (ablation)HILP · ExORL visual URLreturnMixed

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