CounselBench

A large-scale expert evaluation and adversarial benchmarking of LLMs in mental health question answering. Published at ICLR 2026 (Oral).

CounselBench is a comprehensive benchmark for evaluating large language models in mental health counseling question answering. The project includes expert-annotated evaluations and adversarial testing scenarios to rigorously assess LLM capabilities and limitations in sensitive clinical contexts.

Key contributions:

  • Large-scale expert evaluation framework
  • Adversarial benchmarking methodology
  • Comprehensive analysis of LLM strengths and failure modes in mental health QA

Published at ICLR 2026 as an Oral presentation.