Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.13563 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911485623533568 |
|---|---|
| author | Karmore, Aryan |
| author_facet | Karmore, Aryan |
| contents | Linear memory scaling stores $N$ independent expert weight matrices requiring $\mathcal{O}(N \cdot d^2)$ memory, which exceeds edge devices memory budget. Current compression methods like quantization, pruning and low-rank factorization reduce constant factors but leave the scaling bottleneck unresolved. We introduce ButterflyMoE, a method that treats experts not as independent weight matrices but as geometric reorientations of a unified shared quantized substrate. Diversity among experts arises from viewing different angles of shared capacity, not from redundant storage. By applying learned rotations to a shared ternary prototype, each expert yields $\mathcal{O}(d^2 + N \cdot d \log d)$ memory,sub-linear in the number of experts. The key insight: training these rotations with quantization reduces activation outliers and stabilizes extreme low bit training, where static methods collapse. Across language modeling benchmarks, ButterflyMoE achieves 150$\times$ memory reduction at 256 experts with negligible accuracy loss. ButterflyMoE allows multiple experts to fit on edge-constrained devices showing that geometric parameterization breaks linear scaling. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_13563 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ButterflyMoE: Sub-Linear Ternary Experts via Structured Butterfly Orbits Karmore, Aryan Machine Learning Artificial Intelligence Linear memory scaling stores $N$ independent expert weight matrices requiring $\mathcal{O}(N \cdot d^2)$ memory, which exceeds edge devices memory budget. Current compression methods like quantization, pruning and low-rank factorization reduce constant factors but leave the scaling bottleneck unresolved. We introduce ButterflyMoE, a method that treats experts not as independent weight matrices but as geometric reorientations of a unified shared quantized substrate. Diversity among experts arises from viewing different angles of shared capacity, not from redundant storage. By applying learned rotations to a shared ternary prototype, each expert yields $\mathcal{O}(d^2 + N \cdot d \log d)$ memory,sub-linear in the number of experts. The key insight: training these rotations with quantization reduces activation outliers and stabilizes extreme low bit training, where static methods collapse. Across language modeling benchmarks, ButterflyMoE achieves 150$\times$ memory reduction at 256 experts with negligible accuracy loss. ButterflyMoE allows multiple experts to fit on edge-constrained devices showing that geometric parameterization breaks linear scaling. |
| title | ButterflyMoE: Sub-Linear Ternary Experts via Structured Butterfly Orbits |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2601.13563 |