AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Memory, Benchmark & Robots: A Benchmark for Solving Complex Tasks with Reinforcement Learning

Egor Cherepanov, Nikita Kachaev, Alexey Kovalev, Aleksandr Panov

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: Probabilistic and autonomous RL · 37 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
increase action chunk size from K=4 to K=8OpenVLA · MIKASA-Robo ShellGameTouchsuccess rate: 0.12 -> 0.47Improved
increase action chunk size from K=4 to K=8OpenVLA · MIKASA-Robo RememberColor3success rate: 0.21 -> 0.59Improved
increase action chunk size from K=4 to K=8OpenVLA · MIKASA-Robo RememberColor9success rate: 0.08 -> 0.06No clear effect
increase number of colors from 3 to 9Octo-small · MIKASA-Robo RememberColorsuccess rate: 0.45 -> 0.11Hurt
increase number of colors from 3 to 9pi0 (K=4) · MIKASA-Robo RememberColorsuccess rate: 0.35 -> 0.15Hurt
add LSTM memory to PPOPPO-MLP · MIKASA-Robo RememberColor3 (RGB+joints, dense reward)success rate: 25% -> 100%Improved
increase number of cubes from three to five or ninePPO-LSTM · MIKASA-Robo RememberColor (dense reward)success rateHurt
switch from dense to sparse rewardPPO-MLP and PPO-LSTM · MIKASA-Robo RememberColor3success rateHurt
provide full state mode instead of RGB+jointsPPO-MLP · MIKASA-Robo all taskssuccess rate: 25% -> 100%Improved
use RATE (memory transformer) instead of DTRATE vs DT · MIKASA-Robo RememberColor3 (offline RL, dense reward)success rate: 0.05 -> 0.68Improved
use RATE instead of CQLRATE vs CQL · MIKASA-Robo ShellGameTouch (offline RL, dense reward)success rate: 0.02 -> 0.97Improved
use RATE instead of BCRATE vs BC · MIKASA-Robo RememberColor9 (offline RL, dense reward)success rate: 0.15 -> 0.10No clear effect

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