Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

OPPO: Bayesian Value Recursion for Token-Level Credit Assignment in LLM Reasoning

About

Reinforcement learning with verifiable rewards has become the standard recipe for improving LLM reasoning, but the dominant algorithm GRPO assigns a single trajectory-level advantage to every token, diluting the signal at pivotal reasoning steps and injecting noise at uninformative ones. Critic-free alternatives derived from on-policy distillation supply per-token signals through oracle-conditioned likelihood ratios, yet apply each signal in isolation from the trajectory-level evidence accumulated up to that position. We propose Oracle-Prompted Policy Optimization (OPPO), which rests on a single observation: the oracle signal used by prior distillation-style methods for local discrimination is also the natural Bayesian update of the model's belief about eventual success. Accumulating the signal along a trajectory yields, in closed form and at the cost of one extra forward pass, a running estimate of the success probability at every position, together with a token-level advantage that requires no learned value network and no additional rollouts. A first-order analysis factorizes the advantage into the per-token discrimination signal used by distillation methods modulated by a state weight that concentrates credit on genuinely pivotal tokens, with a directional variance-reduction guarantee. The framework admits two estimators differing only in which model scores the evidence: a \textit{self-oracle} that reuses the student and recovers the on-policy distillation reward as a strict special case, and a \textit{teacher-oracle} that delegates scoring to a stronger frozen model. On two base LLMs across seven mathematics, science, and code reasoning benchmarks, OPPO improves over GRPO, DAPO, and SDPO by up to $+6.0$ points on AMC'23 and $+5.2$ points on AIME'24, with gains that widen monotonically with response length.

Yu Li, Rui Miao, Tian Lan, Zhengling Qi• 2026

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningMATH500 (test)--
895
Question AnsweringARC Challenge
Accuracy (ARC)81.6
598
Mathematical ReasoningAIME 2024
Accuracy15
479
Mathematical ReasoningMATH 500
Top-1 Accuracy80.8
384
Mathematical ReasoningAIME 2024 (test)--
209
Mathematical ReasoningGSM8K--
204
Question AnsweringGPQA Diamond
Accuracy33.6
61
Science ReasoningARC-C--
58
Code ReasoningLCB
pass@136.2
26
Mathematical ReasoningAMC 2023
Pass@1 Accuracy61.5
18
Showing 10 of 12 rows

Other info

Follow for update