AI Paper Atlas

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

$V_0$: A Generalist Value Model for Any Policy at State Zero

Yi-Kai Zhang, Zhiyuan Yao, Hongyan Hao, Yueqing Sun, Qi GU, Hui Su, Xunliang Cai, De-Chuan Zhan, Han-Jia Ye

ICML 2026 · Reinforcement learning › Policy gradient and actor-critic · approach: Off-policy actor-critic with expressive policies · 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 in-context policy capability history (V0) instead of coupled value modelvalue model · GRPO tracking, DeepSeek-R1-Distill-Qwen-1.5BIntra AUC (1st epoch): 0.708 -> 0.887Improved
V0 vs Vanilla VM on unseen query IDsvalue model · Strict generalization, 1.5BIntra AUC: 0.560 -> 0.710Improved
use V0-based budget allocation in GRPOQwen3-4B-Instruct-2507 GRPO · OlympiadBench Mean@16accuracy: ~2% increaseImproved
use V0 for inference routingmodel fleet routing · 12 benchmarksaccuracy-cost ParetoImproved
use Residual Dynamic Query connector instead of last_token (ablation)V0 · connector ablationIntra AUC: 0.621 -> 0.705Improved
use Residual Dynamic Query instead of Fixed Query (ablation)V0 · connector ablationIntra AUC: 0.674 -> 0.705Improved
use MLP/Cascaded/MultiScale connectors (ablation)V0 · connector ablationIntra AUC: 0.618 / 0.585 / 0.589 vs 0.705Hurt
add soft CE to pairwise ranking loss (ablation)V0 · loss ablationIntra AUC: 0.578 -> 0.705Improved
add pairwise ranking loss to soft CE (ablation)V0 · loss ablationIntra AUC: 0.686 -> 0.705Improved
add pairwise ranking loss to soft CE (ablation)V0 · loss ablationPairwise accuracy: 0.783 -> 0.839Improved
tune TabPFN head jointly with connector (ablation)V0 · tuning ablationIntra AUC: 0.705 -> 0.594Hurt
add connector over last_token (frozen TabPFN) (ablation)V0 · tuning ablationIntra AUC: 0.621 -> 0.705Improved

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