arXiv:2506. 01969v3 Announce Type: replace-cross Abstract: Efficient inference of Multi-Head Latent Attention (MLA) is challenged by deploying the DeepSeek-R1 671B model on a single Multi-GPU server.
arXiv:2608. 19758v1 Announce Type: new Abstract: Long-context modeling is a pivotal capability for Large Language Models, yet the quadratic complexity of attention remains a critical bottleneck, particularly during the compute-intensive prefilling phase.
arXiv:2607. 22389v1 Announce Type: cross Abstract: With the rapid adoption of long-context large language models (LLMs), the continuously growing KV cache during decoding has become the critical memory bottleneck.
FlashAttention-V is a blocked FlashAttention implementation optimized for scalable vector architectures, designed to reduce the memory bandwidth bottleneck of transformer attention on CPUs. By fusing operations, exploiting parallelism across attention heads, and inter‑head packing, it improves vector register utilization and memory locality, enabling efficient scaling from short to very long vectors. Benchmarks on TinyLlama, Llama 3.2, Qwen2.5, and Pythia‑410M show 22×–42× speedups over scalar FlashAttention in prefill and 8×–11× in decode on a Banana Pi BPI‑F3, while also revealing quantization‑related bottlenecks that limit long‑vector scalability.
The paper introduces MiX, a micro‑inverted‑scaling format that replaces shared exponents with shared mantissas to avoid microscaling collapse in low‑bit vision‑language models. An adaptive dual‑format inference framework (MiX‑MX) maps this format to a custom accelerator, replacing multipliers with shifters. Experiments show 4.5‑bit MiX matches or outperforms NVFP4 accuracy while improving area efficiency by 25 % and delivering 2.3–4.5× speedup with 1.4–2.9× energy savings over the Focus accelerator.
HyQuant introduces a hybrid-precision quantization framework for large language model (LLM) attention modules. It quantizes most attention states to low-bit formats while preserving a small set of vertical‑line tokens and local‑window states in full precision, guided by lightweight attention‑pattern signals. This design achieves near‑lossless accuracy across tasks while improving memory and hardware efficiency.
arXiv:2606. 31519v1 Announce Type: new Abstract: Long-context Large Language Model inference is severely bottlenecked by the massive Key-Value (KV) cache, yet existing sparse attention methods often suffer from static fixed-budget (Top-k) retrieval or rely on proxy scores that are computationally expensive and biased.
arXiv:2609.36760v1 Announce Type: new Abstract: Multi-Head Latent Attention (MLA) enables expressive multi-head attention with compact caches for its content and decoupled RoPE paths, yet cache memor...
The paper introduces a quantization technique for tabular foundation models that focuses on converting queries, keys, and values to FP8 and employing explicit FP8 matrix multiplication to accelerate attention calculations. It emphasizes aligning quantization errors between training and test rows to avoid accuracy loss, and demonstrates up to 1.7× speedup over 16‑bit kernels with no significant accuracy degradation on TabPFN‑v3 and TabICLv2 across TabArena and BeyondArena.
arXiv:2606. 10944v1 Announce Type: new Abstract: We introduce a new tool, Express, for converting a non-causal attention approximation into a causal approximation with matching approximation guarantees.
arXiv:2609.13205v1 Announce Type: cross Abstract: Sparse long-context inference requires efficient token retrieval in both prefill and decode. Existing methods often use different retrieval strategie...
MiX: Micro-Inverted-Scaling for End-to-End Low-Bit Vision-Language Model Acceleration proposes a new quantization format that inverts the traditional microscaling approach by assigning private exponents to each element and a shared mantissa. The adaptive dual-format MiX-MX inference framework maps this format to a custom accelerator, replacing multipliers with shifters. Evaluations show that 4.5-bit MiX matches or surpasses NVFP4 accuracy on multimodal benchmarks while improving area efficiency by 25% and delivering 2.3–4.5× speedup with 1.4–2.9× energy reduction compared to the Focus accelerator.