Adamas is a token‑level sparse attention mechanism that uses the Hadamard transform, bucketization, and 2‑bit compression to create compact representations, followed by lightweight distance‑based top‑k selection for efficient long‑context inference. It achieves near‑lossless performance on Llama‑3.1‑8B‑Instruct and delivers up to 6.5× speedup in self‑attention and 1.7× in end‑to‑end decoding at a 32K context length. Adamas also matches full attention performance on challenging long‑form reasoning benchmarks such as AIME 2024 and 2025.
arXiv:2605. 16928v2 Announce Type: replace-cross Abstract: Long-context inference in large language models is bottlenecked by the quadratic cost of full attention.
arXiv:2607. 19358v1 Announce Type: new Abstract: Recent advances in long chain-of-thought reasoning models such as DeepSeek-R1 have led to increasingly longer inference context lengths under the test-time scaling paradigm.
Prism Attention introduces a proposal‑refined sparse attention framework that replaces static cross‑layer index reuse with a coarse‑to‑fine strategy, allowing layer‑wise adaptability while maintaining efficiency. The method leverages the observation that high‑attention tokens remain largely stable across layers, with changes concentrated near the boundary of previously selected regions. Experiments demonstrate state‑of‑the‑art performance on reasoning and long‑context benchmarks, achieving a 3.4× decoding speedup over FlashAttention at 128K context length.
The paper introduces NAMOH, a native sparse attention mechanism that activates only a subset of heads per token, allowing each head to attend to a limited subsequence of tokens. By scaling the number of heads while keeping the active heads per token fixed, the method shortens head histories and reduces key‑value access without increasing overall storage. Experiments demonstrate that NAMOH can outperform fully activated models with the same parameter count and enable more efficient long‑context inference than smaller dense models.
arXiv:2607. 02980v1 Announce Type: cross Abstract: Scaling modern large language models (LLMs) to long contexts is limited by the quadratic computation cost, and poor length extrapolation of dense attention.
arXiv:2604. 20920v2 Announce Type: replace Abstract: Sparse attention can reduce the cost of long-context inference, but most variants introduce new architectural components.
CompKV introduces a compensation‑aware sparse attention framework for long‑context LLM inference. It partitions tokens into blocks and optimizes token selection to minimize the error introduced by block‑level mean compensation, using compact block‑level statistics. Experiments on RULER and LongBench‑Pro show CompKV outperforms other sparse baselines and achieves up to a 6.85× speedup over full attention.
Elastic Threshold Attention (ETA) is a trainable attention mechanism that dynamically predicts contextual thresholds from query representations, enabling selective pruning of KV cache tokens during long‑context decoding. By multiplicatively suppressing sub‑threshold logits during training, ETA avoids representation collapse and eliminates localized attention sinks, allowing a 1.45B model to match dense attention performance at roughly 85% training sparsity and 38% active decode density. At inference, a custom Triton kernel achieves up to 2.5× faster decoding on sequences up to 512K tokens, and an offline calibration step can further reduce compute by 27% by freezing per‑head thresholds.
arXiv:2610.19210v1 Announce Type: new Abstract: Transformers are efficient to pre-train at scale because attention permits parallel computation across sequence positions, but long-context inference r...
The paper introduces Stratified Sparse Attention (SSA), a framework that blends deterministic top‑k retrieval with score‑guided sampling to reduce the gap between sparse and dense attention in large language models. SSA attends to high‑scoring tokens deterministically while sampling the remaining attention mass using proxy scores, thereby recovering more of the attention distribution. Experiments on benchmarks such as RULER, LongBench, LOFT, and AIME show that SSA outperforms existing sparse methods across models ranging from 3B to 27B parameters, achieving significant accuracy gains and up to 2.96× decode speedup with optimized CUDA kernels.
arXiv:2604. 24432v2 Announce Type: replace-cross Abstract: Long-context ability, has become one of the most important iteration direction of next-generation Large Language Models, particularly in semantic understanding/reasoning, code agentic intelligence and recommendation system.
Block Sparse Flash Attention (BSFA) is a drop‑in replacement for FlashAttention that speeds up long‑context inference by pruning about 50% of computation and memory transfers. It selects the top‑k most important value blocks for each query using exact query‑key similarities and calibrated per‑layer, per‑head thresholds, requiring only a one‑time training‑free calibration. On Llama‑3.1‑8B, BSFA delivers up to 1.13× speedup on LongBench with a 1.1% accuracy drop and up to 1.24× on Needle‑in‑a‑Haystack retrieval with a 1% drop, while the attention kernel itself accelerates by up to 1.38×.