User-turn generation reveals that LLMs' interaction awareness is largely decoupled from task accuracy, remaining near zero in deterministic settings even as accuracy scales to 96.8% on GSM8K.
Simpletom: Exposing the gap between explicit tom inference and implicit tom application in llms.arXiv preprint arXiv:2410.13648
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.AI 2years
2026 2verdicts
UNVERDICTED 2representative citing papers
EnactToM benchmark reveals frontier AI models achieve 0% on functional Theory of Mind task completion in embodied multi-agent settings despite 45% average on literal belief probes.
citing papers explorer
-
Beyond the Assistant Turn: User Turn Generation as a Probe of Interaction Awareness in Language Models
User-turn generation reveals that LLMs' interaction awareness is largely decoupled from task accuracy, remaining near zero in deterministic settings even as accuracy scales to 96.8% on GSM8K.
-
EnactToM: An Evolving Benchmark for Functional Theory of Mind in Embodied Agents
EnactToM benchmark reveals frontier AI models achieve 0% on functional Theory of Mind task completion in embodied multi-agent settings despite 45% average on literal belief probes.