Summary of Opentom: a Comprehensive Benchmark For Evaluating Theory-of-mind Reasoning Capabilities Of Large Language Models, by Hainiu Xu et al.
OpenToM: A Comprehensive Benchmark for Evaluating Theory-of-Mind Reasoning Capabilities of Large Language Models
by Hainiu Xu, Runcong Zhao, Lixing Zhu, Jinhua Du, Yulan He
First submitted to arxiv on: 8 Feb 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Computation and Language (cs.CL)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Neural Theory-of-Mind (N-ToM) benchmark, OpenToM, aims to improve current N-ToM assessments by addressing shortcomings such as ambiguous narratives, lack of personality traits, and limited diversity. The new benchmark features longer, clearer stories; characters with explicit personality traits; actions triggered by character intentions; and questions challenging LLMs’ mental state modeling capabilities. Initial results show that state-of-the-art LLMs excel at tracking physical world mental states but struggle with psychological world mental state tracking. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Neural Theory-of-Mind (N-ToM) is a key skill for machines to understand people’s thoughts and feelings. But current tests have some problems, like using confusing stories or not considering things like personality. To fix this, researchers created OpenToM, a new test that includes longer stories with clear characters, actions driven by intentions, and questions that challenge AI’s understanding of mental states in both the physical and emotional world. The results show that top-performing AI models are good at understanding people’s thoughts about the physical world but struggle to understand their emotions. |
Keywords
» Artificial intelligence » Tracking
