Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention

Vishesh Tripathi posted an arXiv paper on 18 June 2026 (revised 23 June 2026) proposing Grouped Query Experts (GQE), a mixture-of-experts layer built on grouped-query attention (GQA) in which a router selects k query-head experts per token while all key-value heads stay dense. At a 250M parameter scale and a fixed 30B token budget, GQE matched the all-active GQA baseline in downstream accuracy while activating half the query heads per token, preserving GQA's KV cache benefits and cutting only active query-head computation.

Computer Science Machine Learning Submitted on 18 Jun 2026 v1 https://arxiv.org/abs/2606.20945v1 , last revised 23 Jun 2026 this version, v2 Title:Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention View PDF https://arxiv.org/pdf/2606.20945 HTML experimental https://arxiv.org/html/2606.20945v2 Abstract:Self-attention is central to Transformer performance and is often the most expensive part of the Transformer at long context lengths because its pairwise token interactions scale quadratically with sequence length. Standard dense attention also applies the same set of attention heads to every token regardless of token difficulty or information content. This uniform activation can waste compute, especially as sequences grow longer and attention cost increases rapidly. We propose Grouped Query Experts GQE , a mixture-of-experts layer on top of grouped-query attention GQA . Within each GQA group, a router selects k query-head experts per token while all key-value KV heads remain dense and unchanged. Thus, GQE keeps the KV cache benefits of GQA and reduces only the active query-head computation. On a fixed 30B token budget at the 250M parameter scale, GQE matches the all-active GQA baseline in downstream accuracy while activating half the query heads per token. Submission history From: Vishesh Tripathi view email https://arxiv.org/show-email/4d6af19d/2606.20945 Thu, 18 Jun 2026 21:18:42 UTC 3,669 KB \ v1\ https://arxiv.org/abs/2606.20945v1 v2 Tue, 23 Jun 2026 11:48:31 UTC 3,669 KB Additional Features References & Citations Loading... Bibliographic and Citation Tools Bibliographic Explorer What is the Explorer? https://info.arxiv.org/labs/showcase.html arxiv-bibliographic-explorer Connected Papers What is Connected Papers? https://www.connectedpapers.com/about Litmaps What is Litmaps? https://www.litmaps.co/ scite Smart Citations What are Smart Citations? https://www.scite.ai/ Code, Data and Media Associated with this Article alphaXiv What is alphaXiv? https://alphaxiv.org/ CatalyzeX Code Finder for Papers What is CatalyzeX? https://www.catalyzex.com DagsHub What is DagsHub? https://dagshub.com/ Gotit.pub What is GotitPub? http://gotit.pub/faq Hugging Face What is Huggingface? https://huggingface.co/huggingface ScienceCast What is ScienceCast? https://sciencecast.org/welcome Demos Recommenders and Search Tools Influence Flower What are Influence Flowers? https://influencemap.cmlab.dev/ CORE Recommender What is CORE? https://core.ac.uk/services/recommender IArxiv Recommender What is IArxiv? https://iarxiv.org/about arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs https://info.arxiv.org/labs/index.html .