Code metric regression on APPS Leetcode (test)
0.474RMSE+GenRe2-ReMax
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| +GenRe2-ReMaxRL Backbone=ReMax2025.12 | 0.474 | 0.083 | 0.967 | |
| Base Model2025.12 | 0.493 | 0.009 | 0.935 | |
| +CETraining Objective=Cross-Entropy2025.12 | 0.495 | -0.002 | 0.913 | |
| +NTL-WASTraining Objective=Negative Task Likelihood - Wasserstein2025.12 | 0.495 | -0.002 | 0.904 | |
| +NTL-MSETraining Objective=Negative Task Likelihood - Mean Squared Error2025.12 | 0.495 | -0.002 | 0.867 | |
| +DIST2Training Objective=DIST22025.12 | 0.495 | -0.002 | 0.902 |