AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Grounded Reinforcement Learning: Learning to Win the Game under Human Commands

Shusheng Xu, Huaijie Wang, Yi Wu

NeurIPS 2022 · Reinforcement learning › RL from feedback · approach: Theory of RLHF · 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
Replace unconditioned RL training (NA commands) with random-command training (ablation)RED · MiniRTSwin rateHurt
Replace NA training commands with a learned human-proxy command model (ablation)RED · MiniRTSwin rateHurt
Add RED self-distillation over BC-only Switch policy under Oracle commandsRED vs Switch · MiniRTS Oracle commandswin rate (%): 78.7 -> 92.6Improved
Add RED over Joint soft-constraint objective under NA commandsRED vs Joint · MiniRTS NA commandswin rate (%): 51.8 -> 57.8Improved
Add RED over Joint under Oracle commandsRED vs Joint · MiniRTS Oracle commandswin rate (%): 90.2 -> 92.6Improved
Add RED over Joint under Random commandsRED vs Joint · MiniRTS Random commandswin rate (%): 32.3 -> 29.8No clear effect
Add RED over Switch under Random commandsRED vs Switch · MiniRTS Random commandswin rate (%): 12.3 -> 29.8Improved
Add IRL learned reward instead of BC-based constraintIRL · MiniRTS Oracle commandswin rate (%): 89.3 -> 57.3Hurt
Add IRL learned reward on unit building tasks (Dragon)IRL vs RED · MiniRTS Dragon build successsuccess rate (%): 88.8 -> 1.8Hurt
Raise ratio ||Dk||:||D|| from 1:1 to 2:1 (more RL samples) (ablation)RED · MiniRTS Oracle commandswin rate (%): 92.6 -> 87.7Hurt
Lower ratio ||Dk||:||D|| from 1:1 to 1:0.25 (ablation)RED · MiniRTS Oracle commandswin rate (%): 92.6 -> 89.6Hurt
Raise ratio ||Dk||:||D|| from 1:0.25 to 2:1 (ablation)RED · MiniRTS validation NLLNLL: 2.92 -> 3.15Hurt

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