Tree-of-Thoughts Reasoning Quiz

Answers and explanations appear as you take the quiz.

1. How does Tree-of-Thoughts differ from chain-of-thought prompting?
   • It explores multiple branching reasoning paths instead of one linear chain
   • It removes reasoning steps entirely
   • It only works on image inputs
   • It uses no language model at all
2. In Tree-of-Thoughts, what does a single node typically represent?
   • A complete final answer only
   • A partial solution or intermediate 'thought'
   • A user query
   • A trained weight
3. What role does the state evaluator play in Tree-of-Thoughts?
   • It writes the final user-facing answer
   • It deletes the model's memory
   • It scores how promising partial solutions are so weak branches can be pruned
   • It translates between languages
4. On the Game of 24 benchmark, ToT dramatically outperformed chain-of-thought roughly by what jump?
   • From 90% to 95% success
   • No improvement at all
   • From 50% to 51% success
   • From about 4% to about 74% success
5. Which search strategies are commonly used to navigate the thought tree?
   • Breadth-first, depth-first, or beam search
   • Only random guessing
   • Alphabetical sorting
   • Gradient descent on pixels
6. What is the main cost tradeoff of using Tree-of-Thoughts?
   • It reduces accuracy to save time
   • It requires more inference compute in exchange for higher accuracy
   • It needs no model calls beyond one
   • It only works offline
7. At one Tree-of-Thoughts node, what does the thought generator typically produce?
   • A final answer only, with no intermediate states
   • A new model checkpoint
   • Several candidate next thoughts that could extend the partial solution
   • A set of training labels for the prompt
8. What decision does the ToT search procedure make after candidate states are generated and scored?
   • Which tokenizer the model should use
   • Whether to change the model’s weights
   • How to label the original training corpus
   • Which branches to expand and which to prune or stop exploring