Language Modeling on Lambada Standard
14.42PPLSPIRALFORMER-L
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SPIRALFORMER-LModel Scale=1.4B, Topology=MeSH2026.02 | 14.42 | — | |
| MeSHModel Size=Pythia-1.4B, Scheme=Recursive, Layers=4+8R2+4, Variant=+mesh, Parameter reduction percentage=-33.3%2025.10 | 19.39 | — | |
| ConceptLMModel Size=410M2026.02 | 23.2 | 15.31 | |
| ContextLMModel Size=410M2026.02 | 24.95 | 15.82 | |
| MeSHModel Size=Pythia-1B, Scheme=Recursive, Layers=3+5R2+3, Variant=+mesh, Parameter reduction percentage=-31.3%2025.10 | 26.71 | — | |
| Pythia-410MModel Size=410M2026.02 | 31.53 | 17.84 | |
| MeSHModel Size=Pythia-410M, Scheme=Recursive, Layers=4+8R2+4, Variant=+mesh, Parameter reduction percentage=-33.3%2025.10 | 42.51 | — | |
| ContextLMModel Size=1.5B2026.02 | 44.62 | 25.93 | |
| ConceptLMModel Size=1.5B2026.02 | 47.06 | 26.38 | |
| MeSHModel Size=Pythia-410M, Scheme=Recursive, Layers=3+6R3+3, Variant=+mesh, Parameter reduction percentage=-50.0%2025.10 | 52.07 | — | |
| GPT2-1.5BModel Size=1.5B2026.02 | 55.12 | 29.68 | |
| VanillaModel Size=160M, Scheme=Vanilla, Layers=122025.10 | 129.89 | — | |
| MeSHModel Size=160M, Scheme=Recursive, Layers=2+4R2+2, Variant=+mesh, Parameter reduction percentage=-33.3%2025.10 | 178.77 | — |