6 citations · 13 across the 15 of their papers we have counts for
11 papers · 1 filter
DoSA: Dots Sparse Attention
Fanxu Meng, Taosong Fang
Sparse attention visits few keys, yet finding them can still require scoring every key with every indexer head. We introduce DoSA (Dots Sparse Attention), a lossless method for mak…
Beyond KV Reconstruction: Functional Reconstruction for MLA Draft Models in Speculative Decoding
Weiye Shi, Fanxu Meng, Muhan Zhang
Multi-head latent attention (MLA) is increasingly important for long-context LLM inference because compact latent states replace the growing key-value (KV) cache and reduce decodin…
MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference
Ruijie Zhou, Fanxu Meng, Yufei Xu +4
DeepSeek Sparse Attention (DSA) sets the state of the art for fine-grained inference-time sparse attention by introducing a learned token-wise indexer that scores every prefix toke…
GQLA: Group-Query Latent Attention for Hardware-Adaptive Decoding
Fanxu Meng, Zhixin Pan
Decoding efficiency depends on KV-cache traffic, computation, and parallelism. Multi-head Latent Attention (MLA) offers absorbed and head-expanded forms, but its head-specific proj…
HISA: Efficient Hierarchical Indexing for Fine-Grained Sparse Attention
Yufei Xu, Fanxu Meng, Fan Jiang +11
Token-level sparse attention mechanisms, exemplified by DeepSeek Sparse Attention (DSA), achieve fine-grained key selection by scoring every historical key for each query through a…
Breaking the Blocks: Continuous Low-Rank Decomposed Scaling for Unified LLM Quantization and Adaptation
Pingzhi Tang, Ruijie Zhou, Fanxu Meng +2
Current quantization methods for LLMs predominantly rely on block-wise structures to maintain efficiency, often at the cost of representational flexibility. In this work, we demons…
TPLA: Tensor Parallel Latent Attention for Efficient Disaggregated Prefill and Decode Inference
Xiaojuan Tang, Fanxu Meng, Pingzhi Tang +4
Multi-Head Latent Attention (MLA), introduced in DeepSeek-V2, compresses key-value states into a low-rank latent vector, caching only this vector to reduce memory. In tensor parall…
LoRASuite: Efficient LoRA Adaptation Across Large Language Model Upgrades
Yanan Li, Fanxu Meng, Muhan Zhang +3
As Large Language Models (LLMs) are frequently updated, LoRA weights trained on earlier versions quickly become obsolete. The conventional practice of retraining LoRA weights from…
TransMLA: Multi-Head Latent Attention Is All You Need
Fanxu Meng, Pingzhi Tang, Xiaojuan Tang +3
In this paper, we present TransMLA, a framework that seamlessly converts any GQA-based pre-trained model into an MLA-based model. Our approach enables direct compatibility with Dee…
CLOVER: Cross-Layer Orthogonal Vectors Pruning and Fine-Tuning
Fanxu Meng, Pingzhi Tang, Fan jiang +1
Decoder-only models generate tokens autoregressively by caching key/value vectors, but as the cache grows, inference becomes memory-bound. To address this issue, we introduce CLOVE…
PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models
Fanxu Meng, Zhaohui Wang, Muhan Zhang
To parameter-efficiently fine-tune (PEFT) large language models (LLMs), the low-rank adaptation (LoRA) method approximates the model changes throug…