Fig. 4: Invalid actions in graph traversal tasks.
From: A brain-inspired agentic architecture to improve planning with LLMs
`% invalid' indicates the percentage of moves that are invalid (↓ better). GPT-4 Zero-shot, ICL, CoT, and MAD baselines are deterministic, and therefore, a single run was performed on all problems. Note that MAP did not employ tree search on the Steppath task, and did not employ task decomposition on any of the graph traversal tasks. Without tree search, MAP's performance is deterministic, and therefore only a single run was performed on the Steppath task, whereas we performed 5 runs with ToT. Gray error bars reflect 95% binomial confidence intervals (for models evaluated on a single run). Colored dots reflect values of 0%. For Valuepath, Detour, and Reward Revaluation we performed 10, 10, and 5 runs respectively with MAP and ToT, and present average performance ± the standard error of the mean (black error bars).
