# Mining Intrinsic Rewards from LLM Hidden States for Efficient Best-of-N Sampling

arXiv: https://arxiv.org/abs/2505.12225 · HTML: https://mlsanity.com/paper/2505.12225/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Process reward models for reasoning, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 37 of its 76 references that are in the map.
- First posted on arXiv 18 May 2025.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Process reward models for reasoning](https://mlsanity.com/alignment/process-reward-models-for-reasoning/)

## Builds on (most cited first, 30 of 37)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations](https://mlsanity.com/paper/2312.08935/) (2023)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [The Internal State of an LLM Knows When It's Lying](https://mlsanity.com/paper/2304.13734/) (2023)
- [Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problems](https://mlsanity.com/paper/1705.04146/) (2017)
- [DeepSpeed Inference: Enabling Efficient Inference of Transformer Models at Unprecedented Scale](https://mlsanity.com/paper/2207.00032/) (2022)
- [UltraFeedback: Boosting Language Models with Scaled AI Feedback](https://mlsanity.com/paper/2310.01377/) (2023)
- [Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs](https://mlsanity.com/paper/2410.18451/) (2024)
- [PKU-SafeRLHF: Towards Multi-Level Safety Alignment for LLMs with Human Preference](https://mlsanity.com/paper/2406.15513/) (2024)
- [From r to Q^*: Your Language Model is Secretly a Q-Function](https://mlsanity.com/paper/2404.12358/) (2024)
- [AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea](https://mlsanity.com/paper/2411.15738/) (2024)
- [Advancing LLM Reasoning Generalists with Preference Trees](https://mlsanity.com/paper/2404.02078/) (2024)
- [Reasoning Models Know When They're Right: Probing Hidden States for Self-Verification](https://mlsanity.com/paper/2504.05419/) (2025)
- [Can 1B LLM Surpass 405B LLM? Rethinking Compute-Optimal Test-Time Scaling](https://mlsanity.com/paper/2502.06703/) (2025)
- [WebGLM: Towards An Efficient Web-Enhanced Question Answering System with Human Preferences](https://mlsanity.com/paper/2306.07906/) (2023)
- [Noise Contrastive Alignment of Language Models with Explicit Rewards](https://mlsanity.com/paper/2402.05369/) (2024)
- [Search-in-the-Chain: Interactively Enhancing Large Language Models with Search for Knowledge-intensive Tasks](https://mlsanity.com/paper/2304.14732/) (2023)
- [Robust Preference Optimization through Reward Model Distillation](https://mlsanity.com/paper/2405.19316/) (2024)
- [Open Problems in Machine Unlearning for AI Safety](https://mlsanity.com/paper/2501.04952/) (2025)
- [Logits of API-Protected LLMs Leak Proprietary Information](https://mlsanity.com/paper/2403.09539/) (2024)

## Built on by (most cited first, 9 of 9)

- [CLUE: Non-parametric Verification from Experience via Hidden-State Clustering](https://mlsanity.com/paper/2510.01591/) (2025)
- [Sparse Reward Subsystem in Large Language Models](https://mlsanity.com/paper/2602.00986/) (2026)
- [TrajSelector: Harnessing Latent Representations for Efficient and Effective Best-of-N in Large Reasoning Model](https://mlsanity.com/paper/2510.16449/) (2025)
- [Breaking the Solver Bottleneck: Training Task Generators at the Learnable Frontier](https://mlsanity.com/paper/2606.18284/) (2026)
- [A decodability criterion predicts when hidden-state selection beats majority voting in large language models](https://mlsanity.com/paper/2608.17124/) (2026)
- [Graph-Regularized Sparse Autoencoders for LLM Safety Steering](https://mlsanity.com/paper/2512.06655/) (2025)
- [Inverse RL Helps Align AI by Imitating Humans](https://mlsanity.com/paper/2607.24900/) (2026)
- [HSRM: Hidden-State Reward Models for Test-Time Verification](https://mlsanity.com/paper/2608.30841/) (2026)
- [LatentSift: Policy-State Filtering for Token-Efficient Verification of Software Engineering Agents](https://mlsanity.com/paper/2609.36371/) (2026)

## Abstract (arXiv)

Best-of-N sampling is a powerful method for improving Large Language Model (LLM) performance, but it is often limited by its dependence on massive, text-based reward models. These models are not only computationally expensive but also data-hungry, requiring extensive labeled datasets for training. This creates a significant data challenge, as they overlook a rich, readily available data source: the LLM's own internal hidden states. To address this data and efficiency gap, we introduce SWIFT (Simple Weighted Intrinsic Feedback Technique), a novel and lightweight method that learns a reward function directly from the rich information embedded in LLM hidden states. Operating at the token embedding level, SWIFT employs simple linear layers to effectively distinguish between preferred and dispreferred generations, eliminating the need for computationally intensive text-based modeling. Extensive experiments on standard benchmarks show that SWIFT outperforms existing baselines (12.7% higher accuracy than EurusRM-7B on MATH dataset) while using less than 0.005% of their parameters. Its robust scalability, compatibility with certain closed-source models via logit access, and ability to combine with traditional reward models for additional performance highlight SWIFT's practical value and contribution to more efficient data-driven LLM post-training. Our code is available at https://github.com/aster2024/SWIFT .