Metraipolis › archive › dc_48e50e1acccf
TopK-Guided: Adaptive, Budget-Aware Activation Sparsity for Efficient LLM Inference — arXiv:2610.01763
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.01763
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T12:05:59.985Z
- SHA-256 of the archived text
b930b3c5f259c93dfa2acfea583888088c70963dbfd596e23cf0c482cf89af5e- Stable URL of this page
https://metraipolis.com/documento/dc_48e50e1acccf
How to cite
TopK-Guided: Adaptive, Budget-Aware Activation Sparsity for Efficient LLM Inference — arXiv:2610.01763 — arXiv (en), https://arxiv.org/abs/2610.01763 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_48e50e1acccf (Metraipolis archive dc_48e50e1acccf, sha256 b930b3c5f259c93dfa2acfea583888088c70963dbfd596e23cf0c482cf89af5e)
BibTeX
@misc{dc48e50e1acccf,
title = {TopK-Guided: Adaptive, Budget-Aware Activation Sparsity for Efficient LLM Inference — arXiv:2610.01763},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.01763},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_48e50e1acccf; sha256 b930b3c5f259c93dfa2acfea583888088c70963dbfd596e23cf0c482cf89af5e},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.01763. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Activation sparsity speeds up large language model (LLM) inference by setting unimportant activations to zero so that the corresponding computations can be skipped. Existing training-free methods, however, make different trade-offs: threshold-based methods such as TEAL adapt the sparsity level to each token but do not tightly control the realised sparsity, while TopK-based methods such as WINA enforce a fixed sparsity level but use the same sparsity budget for every token. Both also apply the same budget across transformer blocks, despite large differences in block sensitivity. We introduce TopK-Guided, a training-free method that addresses both limitations by combining bounded token-level sparsity adaptation with sensitivity-aware block-level budget allocation. Across Llama-2 and Llama-3 models, TopK-Guided consistently improves perplexity and downstream accuracy over TEAL and WINA while preserving essentially the same sparsitydependent projection compute as WINA, with the largest gains at high sparsity. Ablations show that both components provide complementary improvements. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-10-02 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.01763 — 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 2026-10-06T10:21:27.971Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-10-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).