Advection on Advection
0.283nRMSEORCA (Sequence Doubling)
Evaluation Results
| Method | Links | |
|---|---|---|
| ORCA (Sequence Doubling)Model Family=Pythia, Size=14M, Setup=Sequence Doubling (ours)2025.10 | 0.283 | |
| ORCA (Sequence Doubling)Model Family=Pythia, Size=70M, Setup=Sequence Doubling (ours)2025.10 | 0.302 | |
| ORCA (Sequence Doubling)Model Family=Pythia, Size=160M, Setup=Sequence Doubling (ours)2025.10 | 0.372 | |
| ORCA (Sequence Doubling)Model Family=Pythia, Size=410M, Setup=Sequence Doubling (ours)2025.10 | 0.647 | |
| FPT (Sequence Doubling)Model Family=Pythia, Size=70M, Setup=Sequence Doubling (ours)2025.10 | 0.725 | |
| FPT (Sequence Doubling)Model Family=Pythia, Size=410M, Setup=Sequence Doubling (ours)2025.10 | 0.888 | |
| ORCA (Sequence Doubling)Model Family=Pythia, Size=1.4B, Setup=Sequence Doubling (ours)2025.10 | 1.23 | |
| ORCA (Sequence Doubling)Model Family=Pythia, Size=1B, Setup=Sequence Doubling (ours)2025.10 | 1.27 | |
| FPT (Sequence Doubling)Model Family=Pythia, Size=160M, Setup=Sequence Doubling (ours)2025.10 | 1.58 | |
| FPT (Sequence Doubling)Model Family=Pythia, Size=1.4B, Setup=Sequence Doubling (ours)2025.10 | 1.69 | |
| ORCA (Sequence Doubling)Model Family=GPT-2, Size=137M, Setup=Sequence Doubling (ours)2025.10 | 1.81 | |
| ORCA (Sequence Doubling)Model Family=GPT-2, Size=380M, Setup=Sequence Doubling (ours)2025.10 | 2.04 | |
| FPT (Sequence Doubling)Model Family=Pythia, Size=1B, Setup=Sequence Doubling (ours)2025.10 | 2.04 | |
| ORCA (Sequence Doubling)Model Family=GPT-2, Size=812M, Setup=Sequence Doubling (ours)2025.10 | 2.49 | |
| FPT (Sequence Doubling)Model Family=Pythia, Size=14M, Setup=Sequence Doubling (ours)2025.10 | 2.54 | |
| ORCA (Parallel Flipping)Model Family=Pythia, Size=70M, Setup=Parallel Flipping (ours)2025.10 | 3.98 | |
| ORCA (Parallel Flipping)Model Family=Pythia, Size=14M, Setup=Parallel Flipping (ours)2025.10 | 4.38 | |
| ORCA (Parallel Flipping)Model Family=Pythia, Size=160M, Setup=Parallel Flipping (ours)2025.10 | 5.2 | |
| ORCA (Parallel Flipping)Model Family=Pythia, Size=410M, Setup=Parallel Flipping (ours)2025.10 | 6.43 | |
| ORCA (Parallel Flipping)Model Family=Pythia, Size=1B, Setup=Parallel Flipping (ours)2025.10 | 8.72 | |
| ORCA (Sequence Doubling)Model Family=GPT-2, Size=1.61B, Setup=Sequence Doubling (ours)2025.10 | 8.83 | |
| ORCA (Parallel Flipping)Model Family=Pythia, Size=1.4B, Setup=Parallel Flipping (ours)2025.10 | 9.5 | |
| ORCA (Parallel Flipping)Model Family=GPT-2, Size=137M, Setup=Parallel Flipping (ours)2025.10 | 10.9 | |
| ORCA (Parallel Flipping)Model Family=GPT-2, Size=812M, Setup=Parallel Flipping (ours)2025.10 | 11.9 | |
| ORCA (Parallel Flipping)Model Family=GPT-2, Size=1.61B, Setup=Parallel Flipping (ours)2025.10 | 13.9 | |
| ORCA (Parallel Flipping)Model Family=GPT-2, Size=380M, Setup=Parallel Flipping (ours)2025.10 | 14.4 | |
| FPT (Parallel Flipping)Model Family=Pythia, Size=410M, Setup=Parallel Flipping (ours)2025.10 | 20.7 | |
| FPT (Parallel Flipping)Model Family=Pythia, Size=1B, Setup=Parallel Flipping (ours)2025.10 | 21.6 | |
| FPT (Parallel Flipping)Model Family=Pythia, Size=1.4B, Setup=Parallel Flipping (ours)2025.10 | 21.9 | |
| FPT (Parallel Flipping)Model Family=Pythia, Size=160M, Setup=Parallel Flipping (ours)2025.10 | 22.9 | |
| FPT (Parallel Flipping)Model Family=Pythia, Size=70M, Setup=Parallel Flipping (ours)2025.10 | 23.5 | |
| ORCAModel Family=Pythia, Size=70M, Setup=Original Setup2025.10 | 27.5 | |
| ORCAModel Family=Pythia, Size=160M, Setup=Original Setup2025.10 | 28.6 | |
| ORCAModel Family=Pythia, Size=14M, Setup=Original Setup2025.10 | 28.7 | |
| FPT (Parallel Flipping)Model Family=Pythia, Size=14M, Setup=Parallel Flipping (ours)2025.10 | 29.5 | |
| ORCAModel Family=Pythia, Size=410M, Setup=Original Setup2025.10 | 29.7 | |
| ORCAModel Family=Pythia, Size=1.4B, Setup=Original Setup2025.10 | 31.4 | |
| ORCAModel Family=Pythia, Size=1B, Setup=Original Setup2025.10 | 39.5 | |
| ORCAModel Family=GPT-2, Size=380M, Setup=Original Setup2025.10 | 49.5 | |
| ORCAModel Family=GPT-2, Size=137M, Setup=Original Setup2025.10 | 50.8 | |
| ORCAModel Family=GPT-2, Size=812M, Setup=Original Setup2025.10 | 52.3 | |
| FPT (Parallel Flipping)Model Family=GPT-2, Size=380M, Setup=Parallel Flipping (ours)2025.10 | 53.2 | |
| FPTModel Family=Pythia, Size=1.4B, Setup=Original Setup2025.10 | 53.7 | |
| ORCAModel Family=GPT-2, Size=1.61B, Setup=Original Setup2025.10 | 54.2 | |
| FPTModel Family=Pythia, Size=1B, Setup=Original Setup2025.10 | 54.3 | |
| FPT (Parallel Flipping)Model Family=GPT-2, Size=137M, Setup=Parallel Flipping (ours)2025.10 | 55.1 | |
| FPTModel Family=Pythia, Size=410M, Setup=Original Setup2025.10 | 55.5 | |
| FPTModel Family=Pythia, Size=160M, Setup=Original Setup2025.10 | 59.3 | |
| FPTModel Family=Pythia, Size=70M, Setup=Original Setup2025.10 | 59.7 | |
| FPT (Parallel Flipping)Model Family=GPT-2, Size=1.61B, Setup=Parallel Flipping (ours)2025.10 | 59.8 | |
| FPT (Sequence Doubling)Model Family=GPT-2, Size=137M, Setup=Sequence Doubling (ours)2025.10 | 62.5 | |
| FPT (Parallel Flipping)Model Family=GPT-2, Size=812M, Setup=Parallel Flipping (ours)2025.10 | 62.9 | |
| FPTModel Family=Pythia, Size=14M, Setup=Original Setup2025.10 | 65.2 | |
| FPT (Sequence Doubling)Model Family=GPT-2, Size=380M, Setup=Sequence Doubling (ours)2025.10 | 67.7 | |
| FPTModel Family=GPT-2, Size=137M, Setup=Original Setup2025.10 | 74.8 | |
| FPT (Sequence Doubling)Model Family=GPT-2, Size=812M, Setup=Sequence Doubling (ours)2025.10 | 76.5 | |
| FPT (Sequence Doubling)Model Family=GPT-2, Size=1.61B, Setup=Sequence Doubling (ours)2025.10 | 76.9 | |
| FPTModel Family=GPT-2, Size=380M, Setup=Original Setup2025.10 | 79.7 | |
| FPTModel Family=GPT-2, Size=1.61B, Setup=Original Setup2025.10 | 85.3 | |
| FPTModel Family=GPT-2, Size=812M, Setup=Original Setup2025.10 | 85.4 |