AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Preference Goal Tuning: Post-Training as Latent Control for Frozen Policies

Guangyu Zhao, Kewei Lian, Haoxuan Ru, Borong Zhang, Haowei Lin, Zhancun Mu, Haobo Fu, QIANG FU, Shaofei Cai, Zihao Wang, Yitao Liang

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL fine-tuning of generative models · 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
use DPO instead of behavior cloning on positive trajectories to tune the latent goal (ablation)PGT latent-goal-only · Minecraft task (pretrained 42.0)success rate (%): 18.2 -> 57.2Improved
tune only the latent goal instead of full fine-tuning (in distribution) (ablation)GROOT with DPO · Minecraft SkillForge tasks, IDsuccess rate / reward: 57.2 vs 62.2; 6.58 vs 6.00; 57.8 vs 58.4No clear effect
tune only the latent goal instead of full fine-tuning (out of distribution) (ablation)GROOT with DPO · Minecraft SkillForge tasks, OODperformance: surpasses full fine-tuning by 13.4%Improved
apply PGT to frozen policy versus pretrained policy with expert promptGROOT · Minecraft SkillForge, 17 tasks, IDrelative improvement: +72.0%Improved
apply PGT to frozen policy versus pretrained STEVE-1STEVE-1 · Minecraft SkillForge, 17 tasks, OODrelative improvement: +36.9%Improved
run iterative preference rounds up to three (ablation)PGT · Minecraft tasks ID and OODperformanceImproved
start PGT from any of five initial video prompts including suboptimal ones (ablation)PGT on collect wood · Minecraft collect woodcumulative rewardImproved
use PGT latent goals rather than multi-task learning (out of distribution)GROOT · Minecraft 6-task multi-taskaggregate performanceMixed
add PGT latent goals to GROOT with JARVIS-1 plannerGROOT · long-horizon craftingsuccess rate (%): 37.8 -> 64.9Improved
tune latent goal instead of LoRA, BitFit or VeRA under same preference data (ablation)GROOT · Minecraft ID and OODperformanceImproved
apply PGT to Minecraft tasks where success rate is saturated or zero baselineGROOT/STEVE-1 · Minecraft SkillForge, per-tasksuccess rate: e.g. -65.5%, -80.0%, -0.5% on some tasksMixed

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