# Reward Models Are Secretly Value Functions: Temporally Coherent Reward Modeling

arXiv: https://arxiv.org/abs/2604.22981 · HTML: https://mlsanity.com/paper/2604.22981/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Reward modeling and RLHF, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 22 of its 28 references that are in the map.
- First posted on arXiv 24 Apr 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward modeling and RLHF](https://mlsanity.com/alignment/reward-modeling-and-rlhf/)

## Builds on (most cited first, 22 of 22)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2204.05862/) (2022)
- [The Lessons of Developing Process Reward Models in Mathematical Reasoning](https://mlsanity.com/paper/2501.07301/) (2025)
- [Fine-Grained Human Feedback Gives Better Rewards for Language Model Training](https://mlsanity.com/paper/2306.01693/) (2023)
- [Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs](https://mlsanity.com/paper/2410.18451/) (2024)
- [ProcessBench: Identifying Process Errors in Mathematical Reasoning](https://mlsanity.com/paper/2412.06559/) (2024)
- [From r to Q^*: Your Language Model is Secretly a Q-Function](https://mlsanity.com/paper/2404.12358/) (2024)
- [Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy](https://mlsanity.com/paper/2507.01352/) (2025)
- [Free Process Rewards without Process Labels](https://mlsanity.com/paper/2412.01981/) (2024)
- [RewardBench 2: Advancing Reward Model Evaluation](https://mlsanity.com/paper/2506.01937/) (2025)
- [Dense Reward for Free in Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2402.00782/) (2024)
- [TLCR: Token-Level Continuous Reward for Fine-grained Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2407.16574/) (2024)
- [Improving Large Language Models via Fine-grained Reinforcement Learning with Minimum Editing Constraint](https://mlsanity.com/paper/2401.06081/) (2024)
- [Beyond Imitation: Leveraging Fine-grained Quality Signals for Alignment](https://mlsanity.com/paper/2311.04072/) (2023)
- [Preference-grounded Token-level Guidance for Language Model Fine-tuning](https://mlsanity.com/paper/2306.00398/) (2023)
- [Beyond Sparse Rewards: Enhancing Reinforcement Learning with Language Model Critique in Text Generation](https://mlsanity.com/paper/2401.07382/) (2024)
- [Segmenting Text and Learning Their Rewards for Improved RLHF in Language Model](https://mlsanity.com/paper/2501.02790/) (2025)
- [Incremental Sequence Classification with Temporal Consistency](https://mlsanity.com/paper/2505.16548/) (2025)

## Built on by (most cited first, 1 of 1)

- [The Flip Side of RLHF: On-Policy Feedback for Reward Model Self-Supervised Improvement](https://mlsanity.com/paper/2605.30888/) (2026)

## Abstract (arXiv)

Reward models in RLHF are trained to score only the final token of a response - a choice that discards rich signal from every intermediate position and produces models whose token-level outputs are noise. We argue this is a missed opportunity: a well-trained reward model's output at any token should represent the conditional expectation of the final reward given the response so far. We introduce Temporally Coherent Reward Modeling (TCRM), which induces this property via two regularization terms on top of the standard Bradley-Terry loss, with minimizers provably equal to conditional expectations. The regularizers correspond to Monte Carlo and TD value-learning objectives, establishing a direct connection to RL value functions. TCRM requires zero changes to architecture, data, or inference, yet unlocks three capabilities from one principle: interpretable token-level reward trajectories (middle-token pairwise accuracy improved from 50% to 88.9%, final-token accuracy preserved); state-of-the-art PRM performance on ProcessBench (44.9% average F1) among models trained only on outcome data; and unified reward/value modeling in PPO, reducing peak GPU memory by 27% and step time by 19% with matching LLM quality.