AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

The State of Reinforcement Finetuning for Transformer-based Generative Agents

Shengchao Hu, Peng Wang, Guozheng Ma, Shi Fu, Li Shen, Ya Zhang, Dacheng Tao

ICLR 2026 · Reinforcement learning › Visual and representation RL · approach: Task structure and curricula for meta-RL

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 Q-guided policy improvement term to SFT (QP-SFT) (ablation)SFT, Fullmodel tuning · Meta-RL average (expert)average return: 336.07 -> 372.09Improved
add Q-guided policy improvement term to SFT (QP-SFT) (ablation)SFT, Decorator tuning · Meta-RL average (expert)average return: 321.73 -> 334.16Improved
add Q-guided policy improvement term to SFT (QP-SFT) (ablation)SFT, Adaptor tuning · Meta-RL average (expert)average return: 331.69 -> 335.87Improved
add Q-guided policy improvement term to SFT (QP-SFT) (ablation)SFT, Prompt tuning · Meta-RL average (expert)average return: 315.05 -> 317.07Improved
add Q-guided policy improvement term to DPO (QP-DPO) (ablation)DPO, Fullmodel tuning · Meta-RL average (expert)average return: 326.83 -> 362.13Improved
add Q-guided policy improvement term to DPO (QP-DPO) (ablation)DPO, Decorator tuning · Meta-RL average (expert)average return: 314.66 -> 330.08Improved
add Q-guided policy improvement term to DPO (QP-DPO) (ablation)DPO, Adaptor tuning · Meta-RL average (expert)average return: 333.38 -> 339.46Improved
add Q-guided policy improvement term to DPO (QP-DPO) (ablation)DPO, Prompt tuning · Meta-RL average (expert)average return: 313.73 -> 318.39Improved
add Q-guided term to SFT on MetaWorld with full-model tuning (ablation)SFT · MetaWorld (expert, Fullmodel tuning)return: 441.05 -> 553.36Improved
switch finetuning from Adaptor to Decorator parameters (ablation)PPO · MetaWorldperformance: 3% gainImproved
switch algorithm from PPO to QP-DPO (ablation)PPO with Adaptor · MetaWorldperformance: 10% improvementImproved
increase pretrained agent size with full-model finetuning (12.66M to 49.32M trainable) (ablation)QP-DPO Fullmodel · MetaWorld (medium data)performance: 694.87 -> 375.22Hurt

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