Attention analysis shows that LLM tool selection failures occur at the readout/decision stage, not because the model fails to attend to the correct tool definition.
Spectral guardrails for agents in the wild: Detecting tool use hallucinations via attention topology.arXiv preprint arXiv:2602.08082, 2026
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
OpenHalDet creates a standardized benchmark and open codebase for comparing hallucination detectors across diverse LLM generation scenarios and access settings.
citing papers explorer
-
Looking Is Not Picking: An Attention-Segment Account of Tool-Selection Failures in LLM Agents
Attention analysis shows that LLM tool selection failures occur at the readout/decision stage, not because the model fails to attend to the correct tool definition.