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

Hard

Agents, Tool Use & Product Post-Training5 / 7

Chapter 195 / 7

Agents, Tool Use & Product Post-Training

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

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)

Implement the function/class skeleton 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

⌘/Ctrl↵to submit

AI review

No review yet

Submit your solution for a verdict, a score out of 100 and concrete feedback. Any correct approach passes.

Chapter 195 / 7

Agents, Tool Use & Product Post-Training

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

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)

Implement the function/class skeleton in the editor. Any correct approach is accepted.

Hints0 / 2

Reference solution

Your own code stays in the editor.