Explaining Length Bias in LLM-Based Preference Evaluations

• URL: http://arxiv.org/abs/2407.01085v3
• Date: Sun, 29 Dec 2024 08:52:29 GMT
• Title: Explaining Length Bias in LLM-Based Preference Evaluations
• Authors: Zhengyu Hu, Linxin Song, Jieyu Zhang, Zheyuan Xiao, Tianfu Wang, Zhengyu Chen, Nicholas Jing Yuan, Jianxun Lian, Kaize Ding, Hui Xiong,
• Abstract summary: We decompose the preference evaluation metric, specifically the win rate, into two key components: desirability and information mass.<n>We show that response length impacts evaluations by influencing information mass.<n>We propose AdapAlpaca, a simple yet effective adjustment to win rate measurement.
• Score: 51.07275977870145
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The use of large language models (LLMs) as judges, particularly in preference comparisons, has become widespread, but this reveals a notable bias towards longer responses, undermining the reliability of such evaluations. To better understand such bias, we propose to decompose the preference evaluation metric, specifically the win rate, into two key components: desirability and information mass, where the former is length-independent and related to trustworthiness such as correctness, toxicity, and consistency, and the latter is length-dependent and represents the amount of information in the response. We empirically demonstrated the decomposition through controlled experiments and found that response length impacts evaluations by influencing information mass. To derive a reliable evaluation metric that assesses content quality without being confounded by response length, we propose AdapAlpaca, a simple yet effective adjustment to win rate measurement. Specifically, AdapAlpaca ensures a fair comparison of response quality by aligning the lengths of reference and test model responses under equivalent length intervals.

Related papers

• Towards a Signal Detection Based Measure for Assessing Information Quality of Explainable Recommender Systems [0.5371337604556311]
  We develop an objective metric to evaluate Veracity: the information quality of explanations.<n>To assess the effectiveness of our proposed metric, we set up four cases with varying levels of information quality.
  arXiv  Detail & Related papers  (2025-07-01T20:11:17Z)
• Beyond the Surface: Measuring Self-Preference in LLM Judgments [35.66285592603435]
  Studies show that large language models (LLMs) exhibit self-preference bias when serving as judges.<n>Existing methods typically measure this bias by calculating the difference between the scores a judge model assigns to its own responses and those it assigns to responses from other models.<n>We propose the DBG score, which measures self-preference bias as the difference between the scores assigned by the judge model to its own responses and the corresponding gold judgments.
  arXiv  Detail & Related papers  (2025-06-03T08:12:47Z)
• How Does Response Length Affect Long-Form Factuality [44.91589620660189]
  Despite growing attention to factuality, the effect of response length on factuality remains underexplored.<n>We introduce an automatic and bi-level long-form factuality evaluation framework, which achieves high agreement with human annotations.<n>Using this framework, we find that longer responses exhibit lower factual precision, confirming the presence of length bias.
  arXiv  Detail & Related papers  (2025-05-29T09:47:56Z)
• Pairwise or Pointwise? Evaluating Feedback Protocols for Bias in LLM-Based Evaluation [57.380464382910375]
  We show that the choice of feedback protocol can significantly affect evaluation reliability and induce systematic biases. In particular, we show that pairwise evaluation protocols are more vulnerable to distracted evaluation.
  arXiv  Detail & Related papers  (2025-04-20T19:05:59Z)
• Fostering Appropriate Reliance on Large Language Models: The Role of Explanations, Sources, and Inconsistencies [66.30619782227173]
  Large language models (LLMs) can produce erroneous responses that sound fluent and convincing. We identify several features of LLM responses that shape users' reliance. We find that explanations increase reliance on both correct and incorrect responses. We observe less reliance on incorrect responses when sources are provided or when explanations exhibit inconsistencies.
  arXiv  Detail & Related papers  (2025-02-12T16:35:41Z)
• Uncertainty Estimation of Large Language Models in Medical Question Answering [60.72223137560633]
  Large Language Models (LLMs) show promise for natural language generation in healthcare, but risk hallucinating factually incorrect information. We benchmark popular uncertainty estimation (UE) methods with different model sizes on medical question-answering datasets. Our results show that current approaches generally perform poorly in this domain, highlighting the challenge of UE for medical applications.
  arXiv  Detail & Related papers  (2024-07-11T16:51:33Z)
• Eliminating Biased Length Reliance of Direct Preference Optimization via Down-Sampled KL Divergence [31.03305638930844]
  Direct Preference Optimization (DPO) has emerged as a prominent algorithm for the direct and robust alignment of Large Language Models with human preferences. Despite its promising efficacy, DPO faces a notable drawback: "verbosity" We propose that the issue also stems from an inherent algorithmic length reliance in DPO.
  arXiv  Detail & Related papers  (2024-06-16T14:24:30Z)
• Challenges and Considerations in the Evaluation of Bayesian Causal Discovery [49.0053848090947]
  Representing uncertainty in causal discovery is a crucial component for experimental design, and more broadly, for safe and reliable causal decision making. Unlike non-Bayesian causal discovery, which relies on a single estimated causal graph and model parameters for assessment, causal discovery presents challenges due to the nature of its quantity. No consensus on the most suitable metric for evaluation.
  arXiv  Detail & Related papers  (2024-06-05T12:45:23Z)
• Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators [59.48172585509628]
  We propose a simple regression analysis approach for controlling biases in auto-evaluations. As a real case study, we focus on reducing the length bias of AlpacaEval, a benchmark for chat LLMs. We introduce a length-controlled AlpacaEval that aims to answer the counterfactual question: "What would the preference be if the model's and baseline's output had the same length?"
  arXiv  Detail & Related papers  (2024-04-06T02:29:02Z)
• Mitigating LLM Hallucinations via Conformal Abstention [70.83870602967625]
  We develop a principled procedure for determining when a large language model should abstain from responding in a general domain. We leverage conformal prediction techniques to develop an abstention procedure that benefits from rigorous theoretical guarantees on the hallucination rate (error rate) Experimentally, our resulting conformal abstention method reliably bounds the hallucination rate on various closed-book, open-domain generative question answering datasets.
  arXiv  Detail & Related papers  (2024-04-04T11:32:03Z)
• ODIN: Disentangled Reward Mitigates Hacking in RLHF [127.35607931337019]
  We study the issue of reward hacking on the response length, a challenge emerging in Reinforcement Learning from Human Feedback. A well-formatted, verbose but less helpful response from the LLMs can often deceive LLMs or even human evaluators to achieve high scores. Our approach almost eliminates the reward correlation with length, and improves the obtained policy by a significant margin.
  arXiv  Detail & Related papers  (2024-02-11T22:40:12Z)
• Sample Complexity of Preference-Based Nonparametric Off-Policy Evaluation with Deep Networks [58.469818546042696]
  We study the sample efficiency of OPE with human preference and establish a statistical guarantee for it. By appropriately selecting the size of a ReLU network, we show that one can leverage any low-dimensional manifold structure in the Markov decision process.
  arXiv  Detail & Related papers  (2023-10-16T16:27:06Z)
• Loose lips sink ships: Mitigating Length Bias in Reinforcement Learning from Human Feedback [55.78118035358662]
  Reinforcement learning from human feedback serves as a crucial bridge, aligning large language models with human and societal values. We have identified that the reward model often finds shortcuts to bypass its intended objectives. We propose an innovative solution, applying the Product-of-Experts technique to separate reward modeling from the influence of sequence length.
  arXiv  Detail & Related papers  (2023-10-08T15:14:39Z)
• A Long Way to Go: Investigating Length Correlations in RLHF [59.49656695716066]
  This paper demonstrates, on three diverse settings, that optimizing for response length is a significant factor behind RLHF. We find improvements in reward to largely be driven by increasing response length, instead of other features. Even a purely length-based reward reproduces most downstream RLHF improvements over supervised fine-tuned models.
  arXiv  Detail & Related papers  (2023-10-05T17:38:28Z)
• Human Feedback is not Gold Standard [28.63384327791185]
  We critically analyse the use of human feedback for both training and evaluation. We find that while preference scores have fairly good coverage, they under-represent important aspects like factuality.
  arXiv  Detail & Related papers  (2023-09-28T11:18:20Z)
• Linked shrinkage to improve estimation of interaction effects in regression models [0.0]
  We develop an estimator that adapts well to two-way interaction terms in a regression model. We evaluate the potential of the model for inference, which is notoriously hard for selection strategies. Our models can be very competitive to a more advanced machine learner, like random forest, even for fairly large sample sizes.
  arXiv  Detail & Related papers  (2023-09-25T10:03:39Z)
• Peering Through Preferences: Unraveling Feedback Acquisition for Aligning Large Language Models [32.843361525236965]
  We analyze the effect of sparse feedback on the alignment and evaluation of large language models. We find that preferences from ratings and rankings significantly disagree 60% for both human and AI annotators. Our findings shed light on critical gaps in methods for evaluating the real-world utility of language models.
  arXiv  Detail & Related papers  (2023-08-30T07:35:32Z)
• REAM$\sharp$: An Enhancement Approach to Reference-based Evaluation Metrics for Open-domain Dialog Generation [63.46331073232526]
  We present an enhancement approach to Reference-based EvAluation Metrics for open-domain dialogue systems. A prediction model is designed to estimate the reliability of the given reference set. We show how its predicted results can be helpful to augment the reference set, and thus improve the reliability of the metric.
  arXiv  Detail & Related papers  (2021-05-30T10:04:13Z)
• User and Item-aware Estimation of Review Helpfulness [4.640835690336653]
  We investigate the role of deviations in the properties of reviews as helpfulness determinants. We propose a novel helpfulness estimation model that extends previous ones. Our model is thus an effective tool to select relevant user feedback for decision-making.
  arXiv  Detail & Related papers  (2020-11-20T15:35:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.