Unbiased Evaluation of Large Language Models from a Causal Perspective

Meilin Chen, Jian Tian, Liang Ma, Di Xie, Weijie Chen, Jiang Zhu
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:9105-9120, 2025.

Abstract

Benchmark contamination has become a significant concern in the LLM evaluation community. Previous Agents-as-an-Evaluator address this issue by involving agents in the generation of questions. Despite their success, the biases in Agents-as-an-Evaluator methods remain largely unexplored. In this paper, we present a theoretical formulation of evaluation bias, providing valuable insights into designing unbiased evaluation protocols. Furthermore, we identify two type of bias in Agents-as-an-Evaluator through carefully designed probing tasks on a minimal Agents-as-an-Evaluator setup. To address these issues, we propose the Unbiased Evaluator, an evaluation protocol that delivers a more comprehensive, unbiased, and interpretable assessment of LLMs. Extensive experiments reveal significant room for improvement in current LLMs. Additionally, we demonstrate that the Unbiased Evaluator not only offers strong evidence of benchmark contamination but also provides interpretable evaluation results.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-chen25bi, title = {Unbiased Evaluation of Large Language Models from a Causal Perspective}, author = {Chen, Meilin and Tian, Jian and Ma, Liang and Xie, Di and Chen, Weijie and Zhu, Jiang}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {9105--9120}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/chen25bi/chen25bi.pdf}, url = {https://proceedings.mlr.press/v267/chen25bi.html}, abstract = {Benchmark contamination has become a significant concern in the LLM evaluation community. Previous Agents-as-an-Evaluator address this issue by involving agents in the generation of questions. Despite their success, the biases in Agents-as-an-Evaluator methods remain largely unexplored. In this paper, we present a theoretical formulation of evaluation bias, providing valuable insights into designing unbiased evaluation protocols. Furthermore, we identify two type of bias in Agents-as-an-Evaluator through carefully designed probing tasks on a minimal Agents-as-an-Evaluator setup. To address these issues, we propose the Unbiased Evaluator, an evaluation protocol that delivers a more comprehensive, unbiased, and interpretable assessment of LLMs. Extensive experiments reveal significant room for improvement in current LLMs. Additionally, we demonstrate that the Unbiased Evaluator not only offers strong evidence of benchmark contamination but also provides interpretable evaluation results.} }
Endnote
%0 Conference Paper %T Unbiased Evaluation of Large Language Models from a Causal Perspective %A Meilin Chen %A Jian Tian %A Liang Ma %A Di Xie %A Weijie Chen %A Jiang Zhu %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-chen25bi %I PMLR %P 9105--9120 %U https://proceedings.mlr.press/v267/chen25bi.html %V 267 %X Benchmark contamination has become a significant concern in the LLM evaluation community. Previous Agents-as-an-Evaluator address this issue by involving agents in the generation of questions. Despite their success, the biases in Agents-as-an-Evaluator methods remain largely unexplored. In this paper, we present a theoretical formulation of evaluation bias, providing valuable insights into designing unbiased evaluation protocols. Furthermore, we identify two type of bias in Agents-as-an-Evaluator through carefully designed probing tasks on a minimal Agents-as-an-Evaluator setup. To address these issues, we propose the Unbiased Evaluator, an evaluation protocol that delivers a more comprehensive, unbiased, and interpretable assessment of LLMs. Extensive experiments reveal significant room for improvement in current LLMs. Additionally, we demonstrate that the Unbiased Evaluator not only offers strong evidence of benchmark contamination but also provides interpretable evaluation results.
APA
Chen, M., Tian, J., Ma, L., Xie, D., Chen, W. & Zhu, J.. (2025). Unbiased Evaluation of Large Language Models from a Causal Perspective. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:9105-9120 Available from https://proceedings.mlr.press/v267/chen25bi.html.

Related Material