Results for "lambda".

Explanation Quality Assessment as Ranking with Listwise Rewards

We reformulate explanation quality assessment as a ranking problem rather than a generation problem. Instead of optimizing models to produce a single "best" explanation token-by-token, we train reward…