AbGen: Evaluating Large Language Models in Ablation Study Design and Evaluation for Scientific Research
This addresses the problem of unreliable LLM-based evaluation for complex scientific tasks, specifically in ablation study design for NLP researchers, but is incremental as it builds on existing benchmarking and evaluation methods.
The authors introduced AbGen, a benchmark with 1,500 expert-annotated examples from NLP papers to evaluate LLMs in designing ablation studies, finding a significant performance gap between models like DeepSeek-R1-0528 and human experts in importance, faithfulness, and soundness, and showing that current automated evaluation methods are unreliable for this task.
We introduce AbGen, the first benchmark designed to evaluate the capabilities of LLMs in designing ablation studies for scientific research. AbGen consists of 1,500 expert-annotated examples derived from 807 NLP papers. In this benchmark, LLMs are tasked with generating detailed ablation study designs for a specified module or process based on the given research context. Our evaluation of leading LLMs, such as DeepSeek-R1-0528 and o4-mini, highlights a significant performance gap between these models and human experts in terms of the importance, faithfulness, and soundness of the ablation study designs. Moreover, we demonstrate that current automated evaluation methods are not reliable for our task, as they show a significant discrepancy when compared to human assessment. To better investigate this, we develop AbGen-Eval, a meta-evaluation benchmark designed to assess the reliability of commonly used automated evaluation systems in measuring LLM performance on our task. We investigate various LLM-as-Judge systems on AbGen-Eval, providing insights for future research on developing more effective and reliable LLM-based evaluation systems for complex scientific tasks.