Implement a preference-dataset builder from accepted/rejected suggestions, including a pos

Hard

Agents: Tool Use and Post-Training5 / 7

Chapter 195 / 7

Agents: Tool Use and Post-Training

Hard

Implement a preference-dataset builder from accepted/rejected suggestions, including a position-bias correction.

Target interface

build_preference_dataset(logs, exam_propensity, exam_threshold=0.1)

Complete the starter code in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.

solution.pyPython

Local draft

def build_preference_dataset(logs, exam_propensity, exam_threshold=0.1):
    raise NotImplementedError

Local draft

⌘/Ctrl↵to submit

AI review

Ready when you are

Submit your solution for a score and feedback on correctness, edge cases, and clarity. Any correct approach is accepted.

Chapter 195 / 7

Agents: Tool Use and Post-Training

Hard

Implement a preference-dataset builder from accepted/rejected suggestions, including a position-bias correction.

Target interface

build_preference_dataset(logs, exam_propensity, exam_threshold=0.1)

Complete the starter code in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.