AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Towards Bridging the Gap between Large-Scale Pretraining and Efficient Finetuning for Humanoid Control

Weidong Huang, Zhehan Li, Hangxin Liu, Biao Hou, Yao Su, Jingwen Zhang

ICLR 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Stabilizing deep actor-critic · 5 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 large-scale SAC pretraining before model-based finetuning (ablation)LIFT · Booster T1 finetuning, 1.5 m/stracking of target speedImproved
add world-model pretraining (ablation)LIFT · Booster T1 finetuning, 1.5 m/sfinetuning sample efficiencyImproved
replace physics-informed world model with MBPO ensemble (ablation)LIFT · Booster T1 finetuning, 1.5 m/sepisode returnHurt
raise UTD ratio from 1 to 5-10 in SAC pretraining (ablation)SAC pretraining · T1LowDimJoystickconvergence speedImproved
use larger entropy coefficient alpha in finetuning (ablation)LIFT finetuning · Booster T1, 1.5 m/sconvergence stabilityHurt
use multi-step autoregressive loss horizon (2 or 4) instead of 1 (ablation)world model training · Booster T1, 1.5 m/sstable convergence to target velocityImproved
swap SAC for PPO under deterministic data collection (ablation)SSRL framework preliminary study · Booster T1 target 0.2 m/ssample efficiencyHurt
finetune pretrained policy vs PPO/FastTD3/SAC baselinesBooster T1 sim-to-sim finetuning · Brax velocity trackingconvergenceImproved

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