# Uncertainty Estimation for Language Reward Models

arXiv: https://arxiv.org/abs/2203.07472 · HTML: https://mlsanity.com/paper/2203.07472/

## Key facts

- Cited by 37 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 39 of its 56 references that are in the map.
- First posted on arXiv 14 Mar 2022.

## Placed in

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

## Builds on (most cited first, 30 of 39)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer](https://mlsanity.com/paper/1910.10683/) (2019)
- [Layer Normalization](https://mlsanity.com/paper/1607.06450/) (2016)
- [Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning](https://mlsanity.com/paper/1506.02142/) (2015)
- [On the Opportunities and Risks of Foundation Models](https://mlsanity.com/paper/2108.07258/) (2021)
- [Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles](https://mlsanity.com/paper/1612.01474/) (2016)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Pointer Sentinel Mixture Models](https://mlsanity.com/paper/1609.07843/) (2016)
- [Learning to summarize from human feedback](https://mlsanity.com/paper/2009.01325/) (2020)
- [Fine-Tuning Language Models from Human Preferences](https://mlsanity.com/paper/1909.08593/) (2019)
- [Root Mean Square Layer Normalization](https://mlsanity.com/paper/1910.07467/) (2019)
- [Get To The Point: Summarization with Pointer-Generator Networks](https://mlsanity.com/paper/1704.04368/) (2017)
- [Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift](https://mlsanity.com/paper/1906.02530/) (2019)
- [BitFit: Simple Parameter-efficient Fine-tuning for Transformer-based Masked Language-models](https://mlsanity.com/paper/2106.10199/) (2021)
- [Scaling Language Models: Methods, Analysis & Insights from Training Gopher](https://mlsanity.com/paper/2112.11446/) (2021)
- [Self-Attention with Relative Position Representations](https://mlsanity.com/paper/1803.02155/) (2018)
- [Teaching Machines to Read and Comprehend](https://mlsanity.com/paper/1506.03340/) (2015)
- [HyperNetworks](https://mlsanity.com/paper/1609.09106/) (2016)
- [Aleatoric and Epistemic Uncertainty in Machine Learning: An Introduction to Concepts and Methods](https://mlsanity.com/paper/1910.09457/) (2019)
- [A Survey of Uncertainty in Deep Neural Networks](https://mlsanity.com/paper/2107.03342/) (2021)
- [Weight Uncertainty in Neural Networks](https://mlsanity.com/paper/1505.05424/) (2015)
- [How Can We Know When Language Models Know? On the Calibration of Language Models for Question Answering](https://mlsanity.com/paper/2012.00955/) (2020)
- [A Deep Reinforced Model for Abstractive Summarization](https://mlsanity.com/paper/1705.04304/) (2017)
- [Risk-Constrained Reinforcement Learning with Percentile Risk Criteria](https://mlsanity.com/paper/1512.01629/) (2015)
- [Calibration of Pre-trained Transformers](https://mlsanity.com/paper/2003.07892/) (2020)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [Randomized Prior Functions for Deep Reinforcement Learning](https://mlsanity.com/paper/1806.03335/) (2018)
- [WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation](https://mlsanity.com/paper/2201.05955/) (2022)
- [Want To Reduce Labeling Cost? GPT-3 Can Help](https://mlsanity.com/paper/2108.13487/) (2021)
- [Evaluating Scalable Bayesian Deep Learning Methods for Robust Computer Vision](https://mlsanity.com/paper/1906.01620/) (2019)

## Built on by (most cited first, 30 of 37)

- [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)
- [A Survey of Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2312.14925/) (2023)
- [Reward Model Ensembles Help Mitigate Overoptimization](https://mlsanity.com/paper/2310.02743/) (2023)
- [Helping or Herding? Reward Model Ensembles Mitigate but do not Eliminate Reward Hacking](https://mlsanity.com/paper/2312.09244/) (2023)
- [Routing to the Expert: Efficient Reward-guided Ensemble of Large Language Models](https://mlsanity.com/paper/2311.08692/) (2023)
- [Navigating the Grey Area: How Expressions of Uncertainty and Overconfidence Affect Language Models](https://mlsanity.com/paper/2302.13439/) (2023)
- [Loose lips sink ships: Mitigating Length Bias in Reinforcement Learning from Human Feedback](https://mlsanity.com/paper/2310.05199/) (2023)
- [LoRA ensembles for large language model fine-tuning](https://mlsanity.com/paper/2310.00035/) (2023)
- [Uncertainty-Penalized Reinforcement Learning from Human Feedback with Diverse Reward LoRA Ensembles](https://mlsanity.com/paper/2401.00243/) (2023)
- [Improving Reinforcement Learning from Human Feedback with Efficient Reward Model Ensemble](https://mlsanity.com/paper/2401.16635/) (2024)
- [Bayesian Reward Models for LLM Alignment](https://mlsanity.com/paper/2402.13210/) (2024)
- [GFlowOut: Dropout with Generative Flow Networks](https://mlsanity.com/paper/2210.12928/) (2022)
- [Fine-Tuning Language Models via Epistemic Neural Networks](https://mlsanity.com/paper/2211.01568/) (2022)
- [Gaussian Stochastic Weight Averaging for Bayesian Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2405.03425/) (2024)
- [BayesFormer: Transformer with Uncertainty Estimation](https://mlsanity.com/paper/2206.00826/) (2022)
- [Deep Bayesian Active Learning for Preference Modeling in Large Language Models](https://mlsanity.com/paper/2406.10023/) (2024)
- [Conformal Autoregressive Generation: Beam Search with Coverage Guarantees](https://mlsanity.com/paper/2309.03797/) (2023)
- [Hallucination Detection in LLMs: Fast and Memory-Efficient Fine-Tuned Models](https://mlsanity.com/paper/2409.02976/) (2024)
- [Alignment and Safety in Large Language Models: Safety Mechanisms, Training Paradigms, and Emerging Challenges](https://mlsanity.com/paper/2507.19672/) (2025)
- [On the Robustness of Reward Models for Language Model Alignment](https://mlsanity.com/paper/2505.07271/) (2025)
- [Large Language Models are Miscalibrated In-Context Learners](https://mlsanity.com/paper/2312.13772/) (2023)
- [Making Better Use of Unlabelled Data in Bayesian Active Learning](https://mlsanity.com/paper/2404.17249/) (2024)
- [AssistanceZero: Scalably Solving Assistance Games](https://mlsanity.com/paper/2504.07091/) (2025)
- [Ensembles of Low-Rank Expert Adapters](https://mlsanity.com/paper/2502.00089/) (2025)
- [To Softmax, or not to Softmax: that is the question when applying Active Learning for Transformer Models](https://mlsanity.com/paper/2210.03005/) (2022)
- [FLoRA: Sample-Efficient Preference-based RL via Low-Rank Style Adaptation of Reward Functions](https://mlsanity.com/paper/2504.10002/) (2025)
- [RewardUQ: A Unified Framework for Uncertainty-Aware Reward Models](https://mlsanity.com/paper/2602.24040/) (2026)
- [CoAct: Co-Active LLM Preference Learning with Human-AI Synergy](https://mlsanity.com/paper/2604.17501/) (2026)
- [Distributional Energy-Based Models for Uncertainty-Aware Structured LLM Reasoning](https://mlsanity.com/paper/2605.18871/) (2026)

## Abstract (arXiv)

Language models can learn a range of capabilities from unsupervised training on text corpora. However, to solve a particular problem (such as text summarization) it is typically necessary to fine-tune them on a task-specific dataset. It is often easier for humans to choose between options than to provide labeled data, and prior work has achieved state-of-the-art performance by training a reward model from such preference comparisons. However, collecting a large preference comparison dataset is still expensive -- and the learned reward models are unreliable out-of-distribution. We seek to address these problems via uncertainty estimation, which can improve sample efficiency and robustness using active learning and risk-averse reinforcement learning (RL). Specifically, we use bootstrap aggregating (bagging) to train an ensemble of reward models differing in the initialization of their final layer. Ensembles have proved successful in prior applications of active learning, but we find that in our setting ensemble active learning does not outperform random sampling. Further experiments show that while the aggregate predictions are well-calibrated, the ensemble's estimated epistemic uncertainty is only weakly correlated with model error. We suspect this is because the ensemble members are fine-tuned from a single model and so are similar to one another. This suggests current pre-training methods will need to be modified to support uncertainty estimation, e.g. by training multiple language models.