Metraipolis › archive › dc_da973bdf6e10
Stepped MoE: Segment-Level Routing with Configurable Inference Complexity — arXiv:2610.07348
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.07348
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T11:20:36.897Z
- SHA-256 of the archived text
56c73070ec342e6a516a1720a86a93d1d8a05bfcbe5f9bb5d1a4e2618f23d347- Stable URL of this page
https://metraipolis.com/documento/dc_da973bdf6e10
How to cite
Stepped MoE: Segment-Level Routing with Configurable Inference Complexity — arXiv:2610.07348 — arXiv (en), https://arxiv.org/abs/2610.07348 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_da973bdf6e10 (Metraipolis archive dc_da973bdf6e10, sha256 56c73070ec342e6a516a1720a86a93d1d8a05bfcbe5f9bb5d1a4e2618f23d347)
BibTeX
@misc{dcda973bdf6e10,
title = {Stepped MoE: Segment-Level Routing with Configurable Inference Complexity — arXiv:2610.07348},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.07348},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_da973bdf6e10; sha256 56c73070ec342e6a516a1720a86a93d1d8a05bfcbe5f9bb5d1a4e2618f23d347},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.07348. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Training large language models (LLMs) is resource-intensive, and adapting them for diverse deployment scenarios with varying computational constraints remains challenging. While elastic architectures enable flexible model deployment and sparsely activated models allow input-adaptive computation, existing approaches treat these dimensions independently. Moreover, models catered towards on-device edge inference need to conform to the memory and compute limitations of the serving devices. In this paper, we introduce a unified framework that combines elastic structures with sparsely gated architectures to create models that adapt simultaneously to both deployment constraints and task requirements. Our approach employs a model backbone that conditions on both the context and target efficiency specifications, enabling fine-grained control over the accuracy-efficiency trade-off at inference time. The model learns to activate task-relevant parameters within elastically-nested sub-networks, allowing a single model to span multiple capacity points while maintaining input-adaptive routing. Through experiments we demonstrate that we can create a model that allows the flexibility to use 1,2,3,4 billion parameters while being more accurate than their dense counter-parts (2-5\% on knowledge-intensive benchmarks) and at par with their static versions while delivering similar latency metrics as dense models. Overall, we save on device disk space by sharing the model parameters, allow flexibility of serving based on DRAM and compute available while delivering more accurate results. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.07348 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-07T11:15:24.707Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).