AI Paper Atlas

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

GoldenStart: Q-Guided Priors and Entropy Control for Distilling Flow Policies

He Zhang, Ying Sun, Hui Xiong

ICLR 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Regularized scalable deep value learning · 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
add Q-guided prior and entropy-controlled distillationFQL · OGBench offline averagesuccess rate (%): 38.5 -> 47.1Improved
add Q-guided prior and entropy-controlled distillationFQL · Cube Double Play offlinesuccess rate (%): 36 -> 51.3Improved
add Q-guided prior and entropy-controlled distillationFQL · Puzzle-4x4 offline-to-onlinesuccess rate (%): 38 -> 100Improved
add Q-guided prior and entropy-controlled distillationFQL · OGBench offline-to-online averagesuccess rate (%): 67.6 -> 88.6Improved
add Q-guided prior and entropy-controlled distillationFQL · Visual environments offline averagesuccess rate (%): 65.4 -> 70.9Improved
add Q-guided prior and entropy-controlled distillationFQL · HumanoidMaze Large offlinesuccess rate (%): 7 -> 4.6Hurt
add Q-guided prior and entropy-controlled distillationFQL · Visual Puzzle-3x3 offlinesuccess rate (%): 94 -> 88.67Hurt
add learned Q-guided prior (more candidates 0 -> 5 -> 15) (ablation)GoldenStart · Multi-Crescent onlinereturnImproved
add controllable entropy (vs deterministic w/o CE) (ablation)GoldenStart · Puzzle-4x4 / Multi-Crescent onlinereturnImproved
add Q-guided prior (ablation)GoldenStart (high exploration) · Multi-Crescent onlinereturn: 5.9 -> 14.0Improved
lower target entropy (mult 0.1 -> 0.75) (ablation)GoldenStart · Multi-Crescent onlinereturn: 14.0 -> 9.3Hurt
add rejection sampling (N=32) to FQL (ablation)FQL · Cube-Doublesuccess rate (%): 36 -> 37No clear effect

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