# When Can Proxies Improve the Sample Complexity of Preference Learning?

arXiv: https://arxiv.org/abs/2412.16475 · HTML: https://mlsanity.com/paper/2412.16475/

## 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 27 of its 39 references that are in the map.
- First posted on arXiv 21 Dec 2024.

## Placed in

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

## Builds on (most cited first, 27 of 27)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Scaling Laws for Reward Model Overoptimization](https://mlsanity.com/paper/2210.10760/) (2022)
- [Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2307.15217/) (2023)
- [Learning Complex Dexterous Manipulation with Deep Reinforcement Learning and Demonstrations](https://mlsanity.com/paper/1709.10087/) (2017)
- [RLHF-V: Towards Trustworthy MLLMs via Behavior Alignment from Fine-grained Correctional Human Feedback](https://mlsanity.com/paper/2312.00849/) (2023)
- [A Deep Reinforced Model for Abstractive Summarization](https://mlsanity.com/paper/1705.04304/) (2017)
- [Overcoming Exploration in Reinforcement Learning with Demonstrations](https://mlsanity.com/paper/1709.10089/) (2017)
- [The Effects of Reward Misspecification: Mapping and Mitigating Misaligned Models](https://mlsanity.com/paper/2201.03544/) (2022)
- [A Long Way to Go: Investigating Length Correlations in RLHF](https://mlsanity.com/paper/2310.03716/) (2023)
- [AI Alignment: A Comprehensive Survey](https://mlsanity.com/paper/2310.19852/) (2023)
- [Reward Model Ensembles Help Mitigate Overoptimization](https://mlsanity.com/paper/2310.02743/) (2023)
- [Scaling Laws for Reward Model Overoptimization in Direct Alignment Algorithms](https://mlsanity.com/paper/2406.02900/) (2024)
- [ODIN: Disentangled Reward Mitigates Hacking in RLHF](https://mlsanity.com/paper/2402.07319/) (2024)
- [MaxMin-RLHF: Alignment with Diverse Human Preferences](https://mlsanity.com/paper/2402.08925/) (2024)
- [Reward (Mis)design for Autonomous Driving](https://mlsanity.com/paper/2104.13906/) (2021)
- [Consequences of Misaligned AI](https://mlsanity.com/paper/2102.03896/) (2021)
- [Loose lips sink ships: Mitigating Length Bias in Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2310.05199/) (2023)
- [Iterative Data Smoothing: Mitigating Reward Overfitting and Overoptimization in RLHF](https://mlsanity.com/paper/2401.16335/) (2024)
- [Causal Confusion and Reward Misidentification in Preference-Based Reward Learning](https://mlsanity.com/paper/2204.06601/) (2022)
- [Bayesian Reward Models for LLM Alignment](https://mlsanity.com/paper/2402.13210/) (2024)
- [Meta Discovery: Learning to Discover Novel Classes given Very Limited Data](https://mlsanity.com/paper/2102.04002/) (2021)
- [The Expertise Problem: Learning from Specialized Feedback](https://mlsanity.com/paper/2211.06519/) (2022)
- [Countering Reward Over-optimization in LLM with Demonstration-Guided Reinforcement Learning](https://mlsanity.com/paper/2404.19409/) (2024)
- [Active teacher selection for reward learning](https://mlsanity.com/paper/2310.15288/) (2023)

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

- [Shallow Preference Signals: Large Language Model Aligns Even Better with Truncated Data?](https://mlsanity.com/paper/2505.17122/) (2025)

## Abstract (arXiv)

We address the problem of reward hacking, where maximising a proxy reward does not necessarily increase the true reward. This is a key concern for Large Language Models (LLMs), as they are often fine-tuned on human preferences that may not accurately reflect a true objective. Existing work uses various tricks such as regularisation, tweaks to the reward model, and reward hacking detectors, to limit the influence that such proxy preferences have on a model. Luckily, in many contexts such as medicine, education, and law, a sparse amount of expert data is often available. In these cases, it is often unclear whether the addition of proxy data can improve policy learning. We outline a set of sufficient conditions on proxy feedback that, if satisfied, indicate that proxy data can provably improve the sample complexity of learning the ground truth policy. These conditions can inform the data collection process for specific tasks. The result implies a parameterisation for LLMs that achieves this improved sample complexity. We detail how one can adapt existing architectures to yield this improved sample complexity.