English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Human interactions are deeply rooted in the interplay of thoughts, beliefs,and desires made possible by Theory of Mind (ToM): our cognitive ability tounderstand the mental states of ourselves and others. Although ToM may comenaturally to us, emulating it presents a challenge to even the most advancedLarge Language Models (LLMs). Recent improvements to LLMs' reasoningcapabilities from simple yet effective prompting techniques such asChain-of-Thought have seen limited applicability to ToM. In this paper, we turnto the prominent cognitive science theory "Simulation Theory" to bridge thisgap. We introduce SimToM, a novel two-stage prompting framework inspired bySimulation Theory's notion of perspective-taking. To implement this idea oncurrent ToM benchmarks, SimToM first filters context based on what thecharacter in question knows before answering a question about their mentalstate. Our approach, which requires no additional training and minimalprompt-tuning, shows substantial improvement over existing methods, and ouranalysis reveals the importance of perspective-taking to Theory-of-Mindcapabilities. Our findings suggest perspective-taking as a promising directionfor future research into improving LLMs' ToM capabilities.

Think Twice: Perspective-Taking Improves Large Language Models' Theory-of-Mind Capabilities