AI Paper Atlas

Research map / Reinforcement learning / MDPs and planning

Adaptive Policy Backbone via Shared Network

Bumgeun Park, Donghwan Lee

ICML 2026 · Reinforcement learning › MDPs and planning · approach: Average-reward and regularized MDPs

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
freeze a randomly initialized backbone and train only pre/post linear layers instead of end-to-end RLAPB (random backbone) · Cheetah-vel and Cheetah-diraverage returnNo clear effect
freeze a randomly initialized backbone and train only linear layersAPB (random backbone) · Ant-goalaverage returnImproved
freeze a randomly initialized backbone and train only linear layersAPB (random backbone) · Ant-diraverage returnHurt
adapt both pre- and post-backbone linear layers instead of post-backbone only (linear probing) (ablation)APB (random backbone) · MuJoCo meta-RL tasksaverage returnImproved
use APB with meta-trained frozen backbone instead of fine-tuning-based or context-based meta-RL baselinesAPB · OOD adaptation (Cheetah-vel, Cheetah-vel to Cheetah-dir, Ant-goal, Ant-dir)average returnImproved
use meta-trained frozen backbone instead of randomly initialized frozen backbone in behavior cloning (ablation)APB-BC · BC on OOD tasks with narrow demonstrationsmean episodic returnImproved
freeze a randomly initialized backbone instead of training the full model in behavior cloning (ablation)Random APB-BC · BC on OOD tasks with narrow demonstrationsmean episodic returnImproved
use meta-trained versus random frozen backbone in online RL with extended episode length (ablation)APB · Extended episode length (H_eval > H_train)average returnNo clear effect
freeze backbone and train only linear layers (parameter efficiency)APB · Cheetah-veltrainable parameters and time per iteration: 219,606 -> 6,906; 5.65 ms -> 4.78 msImproved
freeze backbone and train only linear layers (parameter efficiency)APB · Ant-goaltrainable parameters and time per iteration: 247,708 -> 35,008; 5.61 ms -> 4.92 msImproved

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