InterveneBench: Benchmarking LLMs for Intervention Reasoning and Causal Study Design in Real Social Systems
This addresses the need for better benchmarks to assess LLMs' causal reasoning in social science, which is incremental as it builds on existing benchmarking efforts.
The authors tackled the problem of evaluating large language models (LLMs) for intervention reasoning and causal study design in real social systems, by introducing InterveneBench, a benchmark derived from 744 peer-reviewed studies, and found that state-of-the-art LLMs struggle, but their proposed multi-agent framework STRIDES achieved significant performance improvements.
Causal inference in social science relies on end-to-end, intervention-centered research-design reasoning grounded in real-world policy interventions, but current benchmarks fail to evaluate this capability of large language models (LLMs). We present InterveneBench, a benchmark designed to assess such reasoning in realistic social settings. Each instance in InterveneBench is derived from an empirical social science study and requires models to reason about policy interventions and identification assumptions without access to predefined causal graphs or structural equations. InterveneBench comprises 744 peer-reviewed studies across diverse policy domains. Experimental results show that state-of-the-art LLMs struggle under this setting. To address this limitation, we further propose a multi-agent framework, STRIDES. It achieves significant performance improvements over state-of-the-art reasoning models. Our code and data are available at https://github.com/Sii-yuning/STRIDES.