Table 3 PlanBench results

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

Model

Logistics

Mystery BW

GPT-4 Zero-shot

7

0.2

GPT-4 ICL

12

7.8

MAD

16.2

7.3

GPT-4 CoT

17

10.6

MAP

24

27.4

  1. Results reflect % solved problems for a single run.
Back to article page