AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

KAGE-Bench: Fast Known-Axis Visual Generalization Evaluation for Reinforcement Learning

Egor Cherepanov, Daniil Zelezetsky, Aleksandr Panov, Alexey Kovalev

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
shift eval filters (photometric) vs train configPPO-CNN · KAGE-Bench Filterssuccess rate: 0.83 -> 0.11Hurt
shift eval lighting effects vs train configPPO-CNN · KAGE-Bench Effectssuccess rate: 0.82 -> 0.16Hurt
shift eval background vs train configPPO-CNN · KAGE-Bench Backgroundsuccess rate: 0.90 -> 0.42Hurt
shift agent appearance at evalPPO-CNN · KAGE-Bench AgentSR gap (%): 21.1%Hurt
shift distractors at evalPPO-CNN · KAGE-Bench DistractorsSR gap (%): 30.9%Hurt
train on more visual diversity (3 images bg vs one image bg)PPO-CNN · KAGE-Bench BackgroundSR gap (%): 9.6 -> -1.3Improved
add RAD-PPO cutout-color augmentationPPO-CNN · KAGE-Bench 8 hardest configssuccess rateMixed
add crop-resize augmentation, CURL, DrQ, SVEAPPO-CNN · KAGE-Bench hard configsconvergenceHurt
increase same-as-agent distractors from 0 to 11 at evalPPO-CNN · KAGE-Bench Distractorssuccess rateHurt

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