Can LLMs Simulate Personas with Reversed Performance? A Benchmark for Counterfactual Instruction Following
This addresses a limitation in persona simulation for virtual environments, which is incremental as it identifies a specific failure mode in existing LLM capabilities.
The paper tackles the problem that state-of-the-art LLMs cannot simulate personas with reversed performance, such as low-proficiency students, impairing simulation diversity. It proposes the first benchmark for counterfactual instruction following in mathematical reasoning, finding that LLMs, including OpenAI o1, all struggle with this task, with performance worsening when intersecting factors like race are added.
Large Language Models (LLMs) are now increasingly widely used to simulate personas in virtual environments, leveraging their instruction-following capability. However, we discovered that even state-of-the-art LLMs cannot simulate personas with reversed performance (e.g., student personas with low proficiency in educational settings), which impairs the simulation diversity and limits the practical applications of the simulated environments. In this work, using mathematical reasoning as a representative scenario, we propose the first benchmark dataset for evaluating LLMs on simulating personas with reversed performance, a capability that we dub "counterfactual instruction following". We evaluate both open-weight and closed-source LLMs on this task and find that LLMs, including the OpenAI o1 reasoning model, all struggle to follow counterfactual instructions for simulating reversedly performing personas. Intersectionally simulating both the performance level and the race population of a persona worsens the effect even further. These results highlight the challenges of counterfactual instruction following and the need for further research.