Large Language Models (LLMs) incur quadratic attention complexity with input length, creating a major time bottleneck in the prefilling stage. Existing acceleration methods largely exploit attention score sparsity by estimating blocks with high attention scores and applying dynamic sparse attention. In this work, we identify another untapped form of sparsity in the prefilling stage, namely decoding-time contribution sparsity, where many attention blocks exhibit nontrivial attention scores during prefilling yet contribute negligibly to subsequent decoding. Building on this observation, we propose TriangleMix, which replaces dense attention with Triangle attention in a subset of layers. Extensive experiments demonstrate that TriangleMix achieves near-lossless performance on both long-context and long-context reasoning benchmarks, while significantly improving efficiency. For 128K inputs, Triangle attention in the subset of layers achieves a 15.3 × speedup in attention kernel computation, significantly exceeding the acceleration of typical dynamic sparse methods ( 1.9 × to 3.4 × ). Furthermore, TriangleMix can be seamlessly combined with dynamic sparsity approaches, delivering an additional 6%–19% reduction in TTFT over using dynamic sparsity alone.

Long-context language models (LCLMs) can process long context, but still exhibit position bias, also known as “lost in the middle”, which indicates placing key information in the middle of the context will significantly affect performance. To mitigating this, we first explore the micro-level manifestations of position bias, concluding that attention weights are a micro-level expression of position bias. Then we identify that, in addition to position embeddings, positional information in hidden states also contributes to position bias, and it manifests itself in specific channels of hidden states, called positional hidden states. Based on these, we propose a method to mitigate position bias by scaling positional hidden states. Experiments on NaturalQuestions Multi-document QA, KV retrieval and LongBench, using various models including RoPE models, context window-extended models, and Alibi models, demonstrate the effectiveness and generalizability of our approach. Our method can improve performance by up to 15.2% in “lost in the middle” benchmark by modifying just one channel of hidden states. Our code is available at https://aka.ms/PositionalHidden.

Large language models (LLMs) enable long-context tasks but face efficiency challenges due to the growing key-value (KV) cache. We propose LeanK, a learning-based method that prunes unimportant key (K) cache channels by leveraging static channel sparsity. LeanK reduces GPU memory and accelerates decoding without sacrificing accuracy. Experiments demonstrate up to 70% K cache and 16%–18% V cache memory reduction, and 1.45× decoding speedup. We also provide insights into model channels and attention heads during long-context inference by analyzing the learned importance distribution. Our code is anonymously available at https://anonymous.4open.science/r/LeanK-7A87/README.md.