Language Modeling on SST
24.2PPL Change (%)LoRA
Evaluation results
| Method | Links | |
|---|---|---|
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+SST2026.04 | 24.2 | |
| LoRABackbone=Mistral-7B, Learning Rate=10⁻³, Tokens per domain=200K, Training Stage=+HS2026.04 | 18.6 | |
| LoRABackbone=Mistral-7B, Learning Rate=10⁻³, Tokens per domain=200K, Training Stage=+SST2026.04 | 15.4 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+LB2026.04 | 12.3 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+SST2026.04 | 11.9 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+HS2026.04 | 11.7 | |
| GAINBackbone=Mistral-7B, Learning Rate=10⁻³, Tokens per domain=200K, Training Stage=+SST2026.04 | 6.9 | |
| GAINBackbone=Mistral-7B, Learning Rate=10⁻³, Tokens per domain=200K, Training Stage=+HS2026.04 | 6.9 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+MdQ2026.04 | 3.6 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+Fin2026.04 | 3.3 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+Med2026.04 | 2.5 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+WT2026.04 | 2.1 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+PG2026.04 | 1.7 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+Fin2026.04 | 1.6 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+HS2026.04 | 1.6 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+MdQ2026.04 | 1.4 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+WT2026.04 | 1.2 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+LB2026.04 | 0.7 | |
| LoRABackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+PG2026.04 | 0.3 | |
| GAINBackbone=GPT-2 Large (774M), Learning rate=10^-3, Training sequence stage=+Med2026.04 | 0 |