| Challenge: | Format biases in reinforcement learning from human feedback are underexplored . despite its effectiveness, RLHF faces challenges, including policy and regulatory constraints . |
| Approach: | They extend the study of preference biases beyond verbosity bias to a wider range of format biase . they show that with a small amount of biased data, they can inject significant bias into the reward model . |
| Outcome: | The proposed approach can be easily exploited by large language models to achieve higher rankings on popular benchmarks like AlpacaEval and LMSYS Chatbot Arena. |
Similar Papers
LLMs Are Biased Towards Output Formats! Systematically Evaluating and Mitigating Output Format Bias of LLMs (2025.naacl-long)
Copied to clipboard
Do Xuan Long, Ngoc-Hai Nguyen, Tiviatis Sim, Hieu Dao, Shafiq Joty, Kenji Kawaguchi, Nancy F. Chen, Min-Yen Kan
| Challenge: | Using format-following capabilities, state-of-the-art large language models (LLMs) can be leveraged to tailor outputs to specific task formats. |
| Approach: | They propose to define a format bias evaluation metric and establish effective strategies to reduce it. |
| Outcome: | The proposed evaluation reduces the variance in ChatGPT’s performance among wrapping formats from 235.33 to 0.71 (%2) |
Aligning to What? Limits to RLHF Based Alignment (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing studies on RLHF and covert and overt biases in large language models are unclear . et al. analyzed off-the-shelf language models to evaluate their overt and cover racial biase . |
| Approach: | They evaluate the relationship between reinforcement learning from human feedback and biases in large language models. |
| Outcome: | The proposed approach can be used to mitigat covert biases, the authors show . they found that the RLHF approach calcifies model biase . |
Don’t Forget Your Reward Values: Language Model Alignment via Value-based Calibration (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for generating large language models have been criticized for their complexity and instability. |
| Approach: | They propose a value-based calibration method to better align Large Language Models with human preferences. |
| Outcome: | The proposed method surpasses existing methods on AI assistant and summarization datasets, providing impressive generalizability, robustness, and diversity in different settings. |
Not All Voices Are Rewarded Equally: Probing and Repairing Reward Models across Human Diversity (2025.findings-emnlp)
Copied to clipboard
Zihao Li, Feihao Fang, Xitong Zhang, Jiaru Zou, Zhining Liu, Wei Xiong, Ziwei Wu, Baoyu Jing, Jingrui He
| Challenge: | Using real-world datasets, we conduct the most comprehensive study to date, auditing various state-of-the-art reward models across nine sensitive attributes, including age, gender, ethnicity, etc. |
| Approach: | They propose a method to mitigate group disparities in reward modeling by using real-world data. |
| Outcome: | The proposed method is based on a population-based dataset with nine demographic attributes, including gender, ethnicity, age, gender, and ethnicity. |
Beyond Excess and Deficiency: Adaptive Length Bias Mitigation in Reward Models for RLHF (2025.findings-naacl)
Copied to clipboard
| Challenge: | Existing efforts to mitigate length bias in reward models have decreased accuracy . achieving an automatic proxy that perfectly replicates human judgment is challenging in practice. |
| Approach: | They propose an adaptive approach that dynamically adjusts the influence of response length in reward evaluations according to the context of the query. |
| Outcome: | The proposed approach reduces unnecessary verbosity while improving overall response quality. |
Towards Reward Fairness in RLHF: From a Resource Allocation Perspective (2025.acl-long)
Copied to clipboard
| Challenge: | if rewards are imperfect, they can adversely affect the alignment of large language models (LLMs). |
| Approach: | They propose a bias-agnostic method to address the issue of reward unfairness from a resource allocation perspective without specifically designing for each type of bias . they apply methods Fairness Regularization and Fairness Coefficient to achieve fairness in rewards. |
| Outcome: | The proposed method achieves fairness in rewards while minimizing biases . it can be applied to verification and reinforcement learning scenarios . |
Rejected Dialects: Biases Against African American Language in Reward Models (2025.findings-naacl)
Copied to clipboard
| Challenge: | Preference alignment via reward models can introduce new biases, hindering reward models’ fairness and equity. |
| Approach: | They propose a framework for evaluating dialect biases in reward models and conduct a case study on biase . they compare reward models' preferences and behavior on paired White Mainstream English and machine-translated and human-written AAL corpora. |
| Outcome: | The proposed framework evaluates dialect biases in reward models and compares them with paired White Mainstream English (WME) and machine-translated and human-written AAL corpora. |
Constructing Your Model’s Value Distinction: Towards LLM Alignment with Anchor Words Tuning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a study of large language models (LLMs) shows that they can generate outputs that are honest, positive, harmless, etc. |
| Approach: | They propose a method that amplifies logits difference between positive and negative tokens . they propose to use the logits gap to generate positive and positive tokens after alignment . |
| Outcome: | The proposed method achieves effective alignment, but requires fewer computational resources compared to training-time alignment methods. |
The Accuracy Paradox in RLHF: When Better Reward Models Don’t Yield Better Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Reinforcement Learning from Human Feedback (RLHF) significantly enhances Natural Language Processing by aligning language models with human expectations. |
| Approach: | They propose to integrate feedback from humans into RLHF to improve language models by capturing human-like preferences. |
| Outcome: | The proposed model outperforms models trained with moderately accurate reward models on relevance, factuality, and completeness tasks. |
Aligning Large Language Models with Human Preferences through Representation Engineering (2024.acl-long)
Copied to clipboard
Wenhao Liu, Xiaohua Wang, Muling Wu, Tianlong Li, Changze Lv, Zixuan Ling, Zhu JianHao, Cenyuan Zhang, Xiaoqing Zheng, Xuanjing Huang
| Challenge: | Existing methods for achieving this alignment involve employing reinforcement learning from human feedback (RLHF) Existing approaches involve using RLHF to fine-tune LLMs based on human labels . however, RLRF is susceptible to instability during fine- tuning and presents challenges in implementation. |
| Approach: | They propose to use reinforcement learning from human feedback to fine-tune large language models with human preferences to achieve precise control of model behavior. |
| Outcome: | Experiments show that RAHF can be used to capture and manipulate representations to align with a broad spectrum of human preferences or values rather than being confined to a single concept or function. |