Table 5 StrategyQA results

From: A brain-inspired agentic architecture to improve planning with LLMs

Model

Accuracy

ToT

81.7  ± 1.2

GPT-4 CoT

84.7  ± 0.3

MAP

87.7  ± 0.7

Human

87.0

  1. Results reflect accuracy on a fixed (but randomly selected) subset of 100/229 questions averaged over 3 runs ( ± standard error).
Back to article page