Are Bias Evaluation Methods Biased ?

• URL: http://arxiv.org/abs/2506.17111v1
• Date: Fri, 20 Jun 2025 16:11:25 GMT
• Title: Are Bias Evaluation Methods Biased ?
• Abstract summary: The creation of benchmarks to evaluate the safety of Large Language Models is one of the key activities within the trusted AI community.<n>We investigate how robust such benchmarks are by using different approaches to rank a set of representative models for bias and compare how similar are the overall rankings.
• Score: 3.9748528039819977
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: The creation of benchmarks to evaluate the safety of Large Language Models is one of the key activities within the trusted AI community. These benchmarks allow models to be compared for different aspects of safety such as toxicity, bias, harmful behavior etc. Independent benchmarks adopt different approaches with distinct data sets and evaluation methods. We investigate how robust such benchmarks are by using different approaches to rank a set of representative models for bias and compare how similar are the overall rankings. We show that different but widely used bias evaluations methods result in disparate model rankings. We conclude with recommendations for the community in the usage of such benchmarks.

Related papers

• Dynamically Allocating Evaluation Effort for Model Ranking [60.00410111594536]
  We formalize multi-model human evaluation as a best-arm identification problem in a multi-armed bandit setup with correlated arms.<n>We show that it improves discrimination between top-performing models.<n>This makes evaluations faster, cheaper and more aligned with large-scale competition evaluation goals.
  arXiv  Detail & Related papers  (2026-08-04T10:33:04Z)
• To Compare, or Not to Compare: On Methodological Practices in Evaluating Social Bias [61.40435736418359]
  We introduce a unified framework that standardizes heterogeneous benchmarks to contrast isolated demographic assessments with forced-choice comparative settings.<n>We show that comparative settings act as aggressive catalysts for latent discrimination, a shift primarily driven by underspecified contexts.<n>Finally, we demonstrate that this comparative prejudice is a generalized phenomenon that scales positively with model size.
  arXiv  Detail & Related papers  (2026-06-23T13:53:50Z)
• IF-RewardBench: Benchmarking Judge Models for Instruction-Following Evaluation [85.56193980646981]
  We propose IF-RewardBench, a comprehensive meta-evaluation benchmark for instruction-following.<n>For each instruction, we construct a preference graph containing all pairwise preferences among multiple responses.<n>Experiments on IF-RewardBench reveal significant deficiencies in current judge models.
  arXiv  Detail & Related papers  (2026-03-05T02:21:17Z)
• Benchmark^2: Systematic Evaluation of LLM Benchmarks [66.2731798872668]
  We propose Benchmark2, a comprehensive framework comprising three complementary metrics.<n>We conduct experiments across 15 benchmarks spanning mathematics, reasoning, and knowledge domains.<n>Our analysis reveals significant quality variations among existing benchmarks and demonstrates that selective benchmark construction can achieve comparable evaluation performance.
  arXiv  Detail & Related papers  (2026-01-07T14:59:03Z)
• Uncovering Competency Gaps in Large Language Models and Their Benchmarks [11.572508874955659]
  We propose a new method that uses sparse autoencoders (SAEs) to automatically uncover both types of gaps.<n>We found that models consistently underperformed on concepts that stand in contrast to sycophantic behaviors.<n>Our method offers a representation-grounded approach to evaluation, enabling concept-level decomposition of benchmark scores.
  arXiv  Detail & Related papers  (2025-12-06T17:39:47Z)
• BiasFreeBench: a Benchmark for Mitigating Bias in Large Language Model Responses [32.58830706120845]
  Existing studies on bias mitigation methods for large language models (LLMs) use diverse baselines and metrics to evaluate debiasing performance.<n>We introduce BiasFreeBench, an empirical benchmark that comprehensively compares eight mainstream bias mitigation techniques.<n>We will publicly release our benchmark, aiming to establish a unified testbed for bias mitigation research.
  arXiv  Detail & Related papers  (2025-09-30T19:56:54Z)
• The Flaw of Averages: Quantifying Uniformity of Performance on Benchmarks [32.00464870277127]
  We study benchmark reliability from a distributional perspective and introduce benchmark harmony.<n>High harmony is a desirable benchmark property, indicating that the aggregate metric reflects uniform competence across models.<n>By recommending that harmony should be reported alongside accuracy, we reframe evaluation from simple performance averages to a more robust, distributionally reliable measurement of performance.
  arXiv  Detail & Related papers  (2025-09-30T02:14:30Z)
• Where is this coming from? Making groundedness count in the evaluation of Document VQA models [12.951716701565019]
  We argue that common evaluation metrics do not account for the semantic and multimodal groundedness of a model's outputs.<n>We propose a new evaluation methodology that accounts for the groundedness of predictions.<n>Our proposed methodology is parameterized in such a way that users can configure the score according to their preferences.
  arXiv  Detail & Related papers  (2025-03-24T20:14:46Z)
• Rethinking Relation Extraction: Beyond Shortcuts to Generalization with a Debiased Benchmark [53.876493664396506]
  Benchmarks are crucial for evaluating machine learning algorithm performance, facilitating comparison and identifying superior solutions.<n>This paper addresses the issue of entity bias in relation extraction tasks, where models tend to rely on entity mentions rather than context.<n>We propose a debiased relation extraction benchmark DREB that breaks the pseudo-correlation between entity mentions and relation types through entity replacement.<n>To establish a new baseline on DREB, we introduce MixDebias, a debiasing method combining data-level and model training-level techniques.
  arXiv  Detail & Related papers  (2025-01-02T17:01:06Z)
• OffsetBias: Leveraging Debiased Data for Tuning Evaluators [1.5790747258969664]
  We qualitatively identify six types of biases inherent in various judge models. Fine-tuning on our dataset significantly enhances the robustness of judge models against biases.
  arXiv  Detail & Related papers  (2024-07-09T05:16:22Z)
• Backdoor-based Explainable AI Benchmark for High Fidelity Evaluation of Attribution Methods [49.62131719441252]
  Attribution methods compute importance scores for input features to explain the output predictions of deep models. In this work, we first identify a set of fidelity criteria that reliable benchmarks for attribution methods are expected to fulfill. We then introduce a Backdoor-based eXplainable AI benchmark (BackX) that adheres to the desired fidelity criteria.
  arXiv  Detail & Related papers  (2024-05-02T13:48:37Z)
• COBIAS: Assessing the Contextual Reliability of Bias Benchmarks for Language Models [14.594920595573038]
  Large Language Models (LLMs) often inherit biases from the web data they are trained on, which contains stereotypes and prejudices.<n>Current methods for evaluating and mitigating these biases rely on bias-benchmark datasets.<n>We introduce a contextual reliability framework, which evaluates model robustness to biased statements by considering the various contexts in which they may appear.
  arXiv  Detail & Related papers  (2024-02-22T10:46:11Z)
• Don't Make Your LLM an Evaluation Benchmark Cheater [142.24553056600627]
  Large language models(LLMs) have greatly advanced the frontiers of artificial intelligence, attaining remarkable improvement in model capacity. To assess the model performance, a typical approach is to construct evaluation benchmarks for measuring the ability level of LLMs. We discuss the potential risk and impact of inappropriately using evaluation benchmarks and misleadingly interpreting the evaluation results.
  arXiv  Detail & Related papers  (2023-11-03T14:59:54Z)
• Learning Evaluation Models from Large Language Models for Sequence Generation [61.8421748792555]
  We propose a three-stage evaluation model training method that utilizes large language models to generate labeled data for model-based metric development.<n> Experimental results on the SummEval benchmark demonstrate that CSEM can effectively train an evaluation model without human-labeled data.
  arXiv  Detail & Related papers  (2023-08-08T16:41:16Z)
• KPEval: Towards Fine-Grained Semantic-Based Keyphrase Evaluation [69.57018875757622]
  We propose KPEval, a comprehensive evaluation framework consisting of four critical aspects: reference agreement, faithfulness, diversity, and utility. Using KPEval, we re-evaluate 23 keyphrase systems and discover that established model comparison results have blind-spots.
  arXiv  Detail & Related papers  (2023-03-27T17:45:38Z)
• Characterizing Fairness Over the Set of Good Models Under Selective Labels [69.64662540443162]
  We develop a framework for characterizing predictive fairness properties over the set of models that deliver similar overall performance. We provide tractable algorithms to compute the range of attainable group-level predictive disparities. We extend our framework to address the empirically relevant challenge of selectively labelled data.
  arXiv  Detail & Related papers  (2021-01-02T02:11:37Z)
• LOGAN: Local Group Bias Detection by Clustering [86.38331353310114]
  We argue that evaluating bias at the corpus level is not enough for understanding how biases are embedded in a model. We propose LOGAN, a new bias detection technique based on clustering. Experiments on toxicity classification and object classification tasks show that LOGAN identifies bias in a local region.
  arXiv  Detail & Related papers  (2020-10-06T16:42:51Z)
• On the Ambiguity of Rank-Based Evaluation of Entity Alignment or Link Prediction Methods [27.27230441498167]
  We take a closer look at the evaluation of two families of methods for enriching information from knowledge graphs: Link Prediction and Entity Alignment. In particular, we demonstrate that all existing scores can hardly be used to compare results across different datasets. We show that this leads to various problems in the interpretation of results, which may support misleading conclusions.
  arXiv  Detail & Related papers  (2020-02-17T12:26:14Z)

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.