{"ID":2887555,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2508.01261","arxiv_id":"2508.01261","title":"Unifying Mixture of Experts and Multi-Head Latent Attention for Efficient Language Models","abstract":"We present MoE-MLA-RoPE, a novel architecture combination that combines Mixture of Experts (MoE) with Multi-head Latent Attention (MLA) and Rotary Position Embeddings (RoPE) for efficient language modeling. Our approach addresses the fundamental trade-off between model capacity and computational efficiency through three key innovations: (1) fine-grained expert routing with 64 micro-experts and top-$k$ selection, enabling flexible specialization through 3.6 * 10^7 possible expert combinations; (2) shared expert isolation that dedicates 2 always active experts for common patterns while routing to 6 of 62 specialized experts; and (3) gradient-conflict-free load balancing that maintains expert utilization without interfering with primary loss optimization. Extensive experiments on models ranging from 17M to 202M parameters demonstrate that MoE-MLA-RoPE with compression ratio r=d/2 achieves 68% KV cache memory reduction and 3.2x inference speedup while maintaining competitive perplexity (0.8% degradation). Compared to the parameters with 53.9M parameters, MoE-MLA-RoPE improves the validation loss by 6.9% over the vanilla transformers while using 42% fewer active parameters per forward pass. FLOP-matched experiments reveal even larger gains: 11.1% improvement with 3.2x inference acceleration. Automated evaluation using GPT-4 as a judge confirms quality improvements in generation, with higher scores on coherence (8.1/10), creativity (7.9/10) and grammatical correctness (8.2/10). Our results establish that architectural novelty, not parameter scaling, defines the efficiency frontier for resource-constrained language model deployment.","short_abstract":"We present MoE-MLA-RoPE, a novel architecture combination that combines Mixture of Experts (MoE) with Multi-head Latent Attention (MLA) and Rotary Position Embeddings (RoPE) for efficient language modeling. Our approach addresses the fundamental trade-off between model capacity and computational efficiency through thre...","url_abs":"https://arxiv.org/abs/2508.01261","url_pdf":"https://arxiv.org/pdf/2508.01261v1","authors":"[\"Sushant Mehta\",\"Raj Dandekar\",\"Rajat Dandekar\",\"Sreedath Panat\"]","published":"2025-08-02T08:33:30Z","proceeding":"cs.AI","tasks":"[\"cs.AI\"]","methods":"[\"Mixture of Experts\",\"Transformer\",\"Language Model\"]","has_code":false}