ReAct: Detecting the Finish Action
Beginner
The problem, from first principles
The ReAct loop stops when the model emits a Finish action with its answer. Detecting that point tells the runner when to return the final answer rather than call another tool.
From theory to code
Implement first_finish_index(actions), returning where the episode ended, or None.
Constraints
- Match the action name exactly.
Hints
Hint 1
Scan the actions in order and return the first index whose name is Finish.
Keywords
research-papersagentsreasoningtool-use
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.