Training Distribution Generation (no future) on FineWeb sample-10BT
0.373MAUVEσ-temporal
Evaluation Results
| Method | Links | |
|---|---|---|
| σ-temporalrmax=0.4, Backbone=GPT-2 base (124M)2026.06 | 0.373 | |
| GPTrmax=0.4, Backbone=GPT-2 base (124M)2026.06 | 0.355 | |
| AC-GPTrmax=0.4, Backbone=GPT-2 base (124M)2026.06 | 0.337 | |
| σ-randomrmax=0.4, Backbone=GPT-2 base (124M)2026.06 | 0.007 |