Mixture of Weight-shared Heterogeneous Group Attention Experts for Dynamic Token-wise KV Optimization

• URL: http://arxiv.org/abs/2506.13541v1
• Date: Mon, 16 Jun 2025 14:30:17 GMT
• Title: Mixture of Weight-shared Heterogeneous Group Attention Experts for Dynamic Token-wise KV Optimization
• Abstract summary: Transformer models face scalability challenges due to inefficient memory allocation for growing key-value caches.<n>We propose mixSGA, a novel mixture-of-expert approach that dynamically optimize token-wise computation and memory allocation.<n>Our key novelties include: (1) a token-wise expert-choice routing mechanism guided by learned importance scores; (2) weight-sharing across grouped attention projections to minimize parameter overhead; and (3) an auxiliary loss to ensure one-hot routing decisions for training-inference consistency in CLMs.
• Score: 27.994376063677766
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Transformer models face scalability challenges in causal language modeling (CLM) due to inefficient memory allocation for growing key-value (KV) caches, which strains compute and storage resources. Existing methods like Grouped Query Attention (GQA) and token-level KV optimization improve efficiency but rely on rigid resource allocation, often discarding "low-priority" tokens or statically grouping them, failing to address the dynamic spectrum of token importance. We propose mixSGA, a novel mixture-of-expert (MoE) approach that dynamically optimizes token-wise computation and memory allocation. Unlike prior approaches, mixSGA retains all tokens while adaptively routing them to specialized experts with varying KV group sizes, balancing granularity and efficiency. Our key novelties include: (1) a token-wise expert-choice routing mechanism guided by learned importance scores, enabling proportional resource allocation without token discard; (2) weight-sharing across grouped attention projections to minimize parameter overhead; and (3) an auxiliary loss to ensure one-hot routing decisions for training-inference consistency in CLMs. Extensive evaluations across Llama3, TinyLlama, OPT, and Gemma2 model families show mixSGA's superiority over static baselines. On instruction-following and continued pretraining tasks, mixSGA achieves higher ROUGE-L and lower perplexity under the same KV budgets.

Related papers

• GRIP-VLM: Group-Relative Importance Pruning for Efficient Vision-Language Models [19.21570577065149]
  In Vision-Language Models (VLMs) processing a massive number of visual tokens incurs prohibitive computational retraining.<n>GRIP-VLM employs a Group Relative Policy (GRPO) paradigm anchored by supervised pruning to directly explore space selection.
  arXiv  Detail & Related papers  (2026-05-13T11:32:03Z)
• A Replicate-and-Quantize Strategy for Plug-and-Play Load Balancing of Sparse Mixture-of-Experts LLMs [64.8510381475827]
  Sparse Mixture-of-Experts (SMoE) architectures are increasingly used to scale large language models efficiently.<n>SMoE models often suffer from severe load imbalance across experts, where a small subset of experts receives most tokens while others are underutilized.<n>We present a systematic analysis of expert routing during inference and identify three findings: (i) load imbalance persists and worsens with larger batch sizes, (ii) selection frequency does not reliably reflect expert importance, and (iii) overall expert workload and importance can be estimated using a small calibration set.
  arXiv  Detail & Related papers  (2026-02-23T15:11:16Z)
• Fast KVzip: Efficient and Accurate LLM Inference with Gated KV Eviction [50.99402504483692]
  We propose a novel gating-based KV cache eviction method for frozen-weight language models.<n>Our approach integrates seamlessly into both the prefill and decoding stages.<n>Experiments show that our method maintains near-lossless performance while evicting up to 70% of the KV cache.
  arXiv  Detail & Related papers  (2026-01-25T03:07:54Z)
• Low-Rank Key Value Attention [3.7728602841318426]
  Transformer pretraining is increasingly constrained by memory and compute requirements.<n>We propose textitlow-rank KV adaptation (LRKV), a simple modification of multi-head attention.<n>LRKV is a drop-in replacement for standard multi-head attention.
  arXiv  Detail & Related papers  (2026-01-16T17:56:40Z)
• Orchestrating Tokens and Sequences: Dynamic Hybrid Policy Optimization for RLVR [31.43482175098666]
  Reinforcement Learning with Verifiable Rewards (RLVR) offers a promising framework for optimizing large language models in reasoning tasks.<n>Existing RLVR algorithms focus on different granularities, and each has complementary strengths and limitations.<n>We propose Dynamic Hybrid Policy Optimization (DHPO) to bridge GRPO and GSPO within a single clipped surrogate objective.
  arXiv  Detail & Related papers  (2026-01-09T07:57:40Z)
• Training-free Context-adaptive Attention for Efficient Long Context Modeling [57.703159205740185]
  Training-free Context-adaptive Attention (TCA-Attention) is a training-free sparse attention mechanism that selectively attends to only the informative tokens for efficient long-context inference.<n>TCA-Attention achieves a 2.8$times$ speedup and reduces KV cache by 61% at 128K context length while maintaining performance comparable to full attention.
  arXiv  Detail & Related papers  (2025-12-10T01:54:57Z)
• Breaking the MoE LLM Trilemma: Dynamic Expert Clustering with Structured Compression [14.086434595924716]
  Mixture-of-Experts (MoE) Large Language Models (LLMs) face a trilemma of load imbalance, parameter redundancy, and communication overhead.<n>We introduce a unified framework based on dynamic expert clustering and structured compression to address these issues cohesively.
  arXiv  Detail & Related papers  (2025-09-27T10:45:58Z)
• KVCompose: Efficient Structured KV Cache Compression with Composite Tokens [7.922206020386125]
  Large language models (LLMs) rely on key-value (KV) caches for efficient autoregressive decoding.<n>We propose a simple, yet effective, KV cache compression framework based on attention-guided, layer-adaptive composite tokens.<n>Our method achieves significant memory reduction while preserving accuracy, consistently outperforming prior structured and semi-structured methods.
  arXiv  Detail & Related papers  (2025-09-05T14:58:24Z)
• SmallKV: Small Model Assisted Compensation of KV Cache Compression for Efficient LLM Inference [71.20542521694524]
  SmallKV is a small model assisted compensation method for KV cache compression.<n>We show that SmallKV achieves 1.75 - 2.56 times higher throughput than baseline methods.
  arXiv  Detail & Related papers  (2025-08-03T09:15:36Z)
• IGD: Token Decisiveness Modeling via Information Gain in LLMs for Personalized Recommendation [70.2753541780788]
  We introduce an Information Gain-based Decisiveness-aware Token handling (IGD) strategy that integrates token decisiveness into both tuning and decoding.<n>IGD consistently improves recommendation accuracy, achieving significant gains on widely used ranking metrics compared to strong baselines.
  arXiv  Detail & Related papers  (2025-06-16T08:28:19Z)
• KVmix: Gradient-Based Layer Importance-Aware Mixed-Precision Quantization for KV Cache [13.662270631753135]
  Quantization can effectively alleviate the memory pressure caused by KV Cache.<n>We propose a novel mixed-precision quantization method for KV Cache named KVmix.
  arXiv  Detail & Related papers  (2025-05-18T07:04:53Z)
• Token-Efficient RL for LLM Reasoning [0.02488650627593658]
  We propose reinforcement learning strategies tailored for reasoning in large language models (LLMs) under strict memory and compute limits.<n>Building on early policy gradient methods with baseline subtraction, we design critic-free methods that operate on a small, informative subset of output tokens.<n>We show that our methods raise accuracy on the SVAMP benchmark from 46% to over 70% and show strong performance on multi-digit multiplication.
  arXiv  Detail & Related papers  (2025-04-29T14:58:43Z)
• Tactic: Adaptive Sparse Attention with Clustering and Distribution Fitting for Long-Context LLMs [10.52833484759311]
  We propose Tactic, a sparsity-adaptive and calibration-free sparse attention mechanism.<n>It dynamically selects tokens based on their cumulative attention scores rather than a fixed token budget.<n>We show that Tactic outperforms existing sparse attention algorithms, achieving superior accuracy and up to 7.29x decode attention speedup.
  arXiv  Detail & Related papers  (2025-02-17T08:39:43Z)
• HAFLQ: Heterogeneous Adaptive Federated LoRA Fine-tuned LLM with Quantization [55.972018549438964]
  Federated fine-tuning of pre-trained Large Language Models (LLMs) enables task-specific adaptation across diverse datasets while preserving privacy.<n>We propose HAFLQ (Heterogeneous Adaptive Federated Low-Rank Adaptation Fine-tuned LLM with Quantization), a novel framework for efficient and scalable fine-tuning of LLMs in heterogeneous environments.<n> Experimental results on the text classification task demonstrate that HAFLQ reduces memory usage by 31%, lowers communication cost by 49%, improves accuracy by 50%, and achieves faster convergence compared to the baseline method.
  arXiv  Detail & Related papers  (2024-11-10T19:59:54Z)
• Mixture Compressor for Mixture-of-Experts LLMs Gains More [71.0473038084673]
  We propose a training-free Mixture-Compressor for Mixture-of-Experts large language models (MoE-LLMs)<n>Our MC integrates static quantization and dynamic pruning to collaboratively achieve extreme compression for MoE-LLMs with less accuracy loss.<n>For instance, at 2.54 bits, MC compresses 76.6% of the model, with only a 3.8% average accuracy loss.
  arXiv  Detail & Related papers  (2024-10-08T18:09:38Z)
• Self-Supervised Graph Embedding Clustering [70.36328717683297]
  K-means one-step dimensionality reduction clustering method has made some progress in addressing the curse of dimensionality in clustering tasks. We propose a unified framework that integrates manifold learning with K-means, resulting in the self-supervised graph embedding framework.
  arXiv  Detail & Related papers  (2024-09-24T08:59:51Z)
• Model Tells You Where to Merge: Adaptive KV Cache Merging for LLMs on Long-Context Tasks [21.815661269986425]
  We propose a novel KV cache merging approach, called KVMerger, to achieve adaptive KV cache compression for long-context tasks. Our approach is inspired by the intriguing observation that key states exhibit high similarity at the token level within a single sequence. We conduct extensive experiments to demonstrate the effectiveness of KVMerger for long-context tasks under constrained memory budgets.
  arXiv  Detail & Related papers  (2024-07-11T12:50:42Z)
• Effectively Compress KV Heads for LLM [28.0801697946958]
  We propose a novel approach for compressing Key-Value ( KV) caches. Our method can compress half or even three-quarters of KV heads while maintaining performance comparable to the original LLMs.
  arXiv  Detail & Related papers  (2024-06-11T08:37:33Z)
• LoRA-Ensemble: Efficient Uncertainty Modelling for Self-Attention Networks [52.46420522934253]
  We introduce LoRA-Ensemble, a parameter-efficient ensembling method for self-attention networks.<n>The method not only outperforms state-of-the-art implicit techniques like BatchEnsemble, but even matches or exceeds the accuracy of an Explicit Ensemble.
  arXiv  Detail & Related papers  (2024-05-23T11:10:32Z)
• Semantic Equitable Clustering: A Simple and Effective Strategy for Clustering Vision Tokens [57.37893387775829]
  We introduce a fast and balanced clustering method, named textbfSemantic textbfEquitable textbfClustering (SEC) SEC clusters tokens based on their global semantic relevance in an efficient, straightforward manner. We propose a versatile vision backbone, SECViT, to serve as a vision language connector.
  arXiv  Detail & Related papers  (2024-05-22T04:49:00Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.