Tensor TownISSUE #65 of 120

tree of thoughts · branching reasoning

NeuraVSThe Overfit Ogre

Neura saysTree of Thoughts explores multiple reasoning branches instead of one linear chain.

Where chain-of-thought is a single path, Tree of Thoughts generates several candidate reasoning steps, evaluates them, and explores the promising branches (like search) — backtracking from dead ends. It improves hard problems (planning, puzzles) at the cost of many more model calls. It's one of several "inference-time search" techniques that trade extra compute for better answers.

Power-ups you unlock

Explore multiple reasoning branches
Evaluate and expand promising paths
Backtrack from dead ends (search)
More compute → better hard-problem answers

The Overfit Ogre attacks — common mistakes

Using it for simple tasks (wasteful)
Ignoring the multiplied token/compute cost
No branch-evaluation heuristic

Boss battleContrast chain-of-thought (one path) with tree-of-thoughts (search).

Example code

<!doctype html><html><head><meta charset="utf-8"></head>
<body style="background:#06040d;color:#e6e0ff;font-family:monospace;padding:20px"><pre>CoT:  step → step → answer (one path)
ToT:  branch → evaluate → expand best → backtrack
costs more, solves harder problems</pre></body></html>

▶ Open the interactive comic issue

‹ Chain Of Thought · Think Before Answering React · Reason + Act Loops ›