Hierarchical Attention

The quadratic scaling of attention is a fundamental bottleneck limiting the long-context capabilities of large language models (LLMs). A variety of approaches have attempted to circumvent this, but they either fail to match attention's expressiveness (e.g., sliding-window or linear attention variants with fixed-size state) or only achieve constant-factor speedups (e.g., sparse attention methods whose indexers themselves scale quadratically). In this paper, we derive an asymptotically subquadratic sparse attention method based on a hierarchical, tree-structured indexer, which we call Zyphra Hierarchical Attention (ZHA). ZHA constructs a tree of recursively compressed key summaries over the KV cache, which future queries traverse from the top down to retrieve a sparse subset of keys. Crucially, by iteratively growing the tree's depth, the total cost including routing and final attention grows sublinearly in sequence length, resulting in overall subquadratic (but superlinear) scaling for attention. We demonstrate that ZHA can be cheaply applied to existing pretrained models with only a short KL distillation training stage over a few billion tokens, and maintains performance on long-context retrieval and challenging reasoning, mathematical, and agentic coding evaluation tasks. Moreover, we show that our method is highly robust to tree geometry including very deep trees of up to 6 levels. Finally, we develop a detailed performance model and show that our approach scales subquadratically, dramatically outperforms competing methods beyond 1 million context, and ultimately has the potential to extend effective context windows to the hundreds of millions or billions of tokens.

Robotic agents operating in real-world environments must adapt continuously from a stream of multimodal demonstrations, often under constraints that preclude storing or revisiting past data. In this replay-free, single-pass lifelong imitation learning setting, sequential updates can progressively distort the representation space, making expert-based approaches particularly sensitive to routing interference, causing expert selection to degrade over time. We propose HIDRA, a structured routing framework that mitigates this issue through a hierarchical dual-attention mechanism, which decouples instruction-level expert retrieval from context-dependent refinement. To further stabilize routing under representation drift, HIDRA introduces a key-level geometric regularization that enforces both alignment within tasks and separation across tasks. The resulting approach enables reliable expert selection while supporting an expandable set of experts for continual adaptation. We evaluate HIDRA on several LIBERO benchmark suites, including more challenging variants with heterogeneous task sequences and paraphrased language instructions. Our approach consistently outperforms both replay-free and replay-based baselines, improving AUC and forward transfer while maintaining low forgetting, with stronger gains under high interference and language variation. The code is available at: https://github.com/yfqi/HIDRA.

Parameter-efficient fine-tuning reduces model and optimizer memory, but dense attention still makes long training sequences expensive. We combine Hierarchical Global Attention (HGA) with segment-wise backpropagation and tiered KV storage. Only the active segment remains differentiable in VRAM; older KV is detached into RAM or NVMe, and HGA loads a bounded set of exact historical tokens for each query block. On Qwen3-8B with 4-bit QLoRA and PG19, dense training on a 16 GB Quadro RTX 5000 fits 2,048 tokens but fails at 4,096, whereas HGA reaches 16,384 tokens with 15.28 GB peak VRAM. Under evaluation the same adapter runs through 131,072 tokens on this card; VRAM is not constant but grows gently with the resident chunk summaries, so RAM and NVMe capacity set the practical limit beyond these lengths. At the shared 2K training length, HGA-trained and dense-trained adapters obtain 2.7405 and 2.7383 nat under the same dense-attention readout, while the stock model obtains 2.9541. At this boundary HGA training is already marginally faster (217.75 vs. 207.02 tokens/s), and the HGA-to-dense throughput ratio improves from 1K to 2K; because HGA keeps the attended historical set per token approximately constant while dense work per token grows, we expect this lead to widen as context grows. Dense attention is used for the main quality and retrieval comparisons so that they measure the learned weights and remain compatible with standard generation frameworks. HGA can also be used for retrieval and generation; an optimized production-grade serving implementation is under development.

Scaling modern large language models (LLMs) to long contexts is limited by the quadratic computation cost, and poor length extrapolation of dense attention. Chunk-wise sparse attention offers a promising alternative, but all existing methods fall short of full attention because of their inaccurate chunk selection. We propose Hierarchical Landmark Sparse (HiLS) Attention, a chunk-wise sparse attention mechanism that learns chunk selection end-to-end under the language-modeling (LM) loss. HiLS factorizes attention hierarchically: each query performs attention independently with each retrieved chunk to extract chunk-specific information, and the resulting outputs are fused according to chunk retrieval scores. By incorporating retrieval scores into the forward attention computation, HiLS optimizes them directly with the LM loss, enabling end-to-end retrieval learning and native sparse training. Experimental results show that HiLS-Attention achieves performance comparable to, and in some cases better than, full attention at in-domain context lengths. Meanwhile, HiLS-Attention extrapolates more than $64\times$ the training context length with 90% retrieval accuracy, far beyond full attention. Moreover, existing full-attention models can be converted to HiLS-Attention with lightweight continued pretraining, preserving in-domain performance while acquiring ultra-long-context extrapolation. Together with its sparse KV access and computation, HiLS-Attention breaks the usual efficiency-performance trade-off, enabling long-context LLMs that are both more efficient and more effective on general long-context tasks than their full-attention counterparts.

Hierarchical Global Attention (HGA) is a drop-in replacement for dense causal attention in pretrained long-context transformers. HGA preserves the original checkpoint parameters: the pretrained $W_Q$, $W_K$, $W_V$, and $W_O$ projections remain unchanged, no calibration parameters are introduced, and no retraining is required. Applied to Qwen3-30B-A3B-Instruct-2507-FP8 on a single RTX~5090 (32GB), the patched model runs out of the box at a 64K-token context, where token-level K/V storage is not feasible on this hardware. Unlike previous sparse-attention methods, HGA performs hierarchical two-level routing. It first retrieves relevant chunks using compact RoPE-aware summaries and then refines the selection by routing only the most relevant groups before performing exact token-level attention. This hierarchical retrieval significantly reduces the number of fetched tokens while preserving exact attention over the retrieved token set, making RAM- and NVMe-backed storage practical. The full historical token K/V resides in host RAM or NVMe storage, while only a small routed working set is transferred to GPU memory during attention. Consequently, GPU memory consumption depends primarily on model weights and the routed working set rather than on the total context length. Across all tested context lengths (4K - 64K tokens), routed attention remains within approximately $0.01$--$0.02$ nats of dense attention while the sparsity used is just about 3%. These results suggest that the approximation introduced by hierarchical routing is small, and that the remaining quality gap is likely dominated by long-context positional encoding rather than by the routing algorithm itself.

Real-time jet tagging is critical for identifying short-lived particle decays in the high-throughput detectors of the Large Hadron Collider, where real-time trigger systems responsible for deciding which collision events to store impose strict latency and accuracy constraints. While transformer architectures achieve the highest jet tagging accuracy when compute is unconstrained, their quadratic self-attention cost makes inference restrictive on trigger budget. Existing efficient variants reduce the computational cost, but hinder the classification performance. To address this limitation, we introduce the Patch Hierarchical Attention Transformer (PHAT-JeT), which combines two mechanisms: a physics-inspired geometric message-passing module that encodes local detector-plane structure, and a hierarchical patch-based attention scheme that computes exact attention within small particle groups while preserving global context through lightweight patch-token communication. Within a restricted budget, PHAT-JeT achieves state-of-the-art accuracy and background rejection among all resource-constrained jet tagging models on four benchmarks (hls4ml, JetClass, Top Tagging, and Quark--Gluon). Our code is available at https://github.com/aaronw5/PHAT-JeT.

Current hierarchical attention methods, such as NSA and InfLLMv2, select the top-k relevant key-value (KV) blocks based on coarse attention scores and subsequently apply fine-grained softmax attention on the selected tokens. However, the top-k operation assumes the number of relevant tokens for any query is fixed and it precludes the gradient flow between the sparse and dense stages. In this work, we propose DashAttention (Differentiable and Adaptive Sparse Hierarchical Attention), which leverages the adaptively sparse $α$-entmax transformation to select a variable number of blocks according to the current query in the first stage. This in turn provides a prior for the second-stage softmax attention, keeping the entire hierarchy fully differentiable. Contrary to other hierarchical attention methods, we show that DashAttention is non-dispersive, translating to better long-context modeling ability. Experiments with large language models (LLMs) show that DashAttention achieves comparable accuracy as full attention with 75% sparsity and a better Pareto frontier than NSA and InfLLMv2, especially in high-sparsity regimes. We also provide an efficient, GPU-aware implementation of DashAttention in Triton, which achieves a speedup of up to over FlashAttention-3 at inference time. Overall, DashAttention offers a cost-effective strategy to model long contexts.

Training causal transformers at extreme sequence lengths is bottlenecked by the quadratic time and memory of scaled dot-product attention (SDPA). In this work, we propose Lighthouse Attention, a training-only symmetrical selection-based hierarchical attention algorithm that wraps around ordinary SDPA and can be easily removed towards the end of the training. Our hierarchical selection is also gradient-free, which exempts us from dealing with a complicated and potentially inefficient backward pass kernel. Our contribution is three-fold: (i) A subquadratic hierarchical pre- and post-processing step that does adaptive compression and decompression of the sequence. (ii) A symmetrical compression strategy that pools queries, keys and values at the same time, while preserving left-to-right causality, which greatly improves parallelism. (iii) A two stage training approach which we pre-train for the majority of the time with Lighthouse Attention and recover a full attention model at the end with a short training. We run preliminary small scale LLM pre-training experiments that show the effectiveness of our method compared to full attention training with all other settings matched, where we achieve a faster total training time and lower final loss after the recovery phase. Full code is available at: https://github.com/ighoshsubho/lighthouse-attention

Image-to-point-cloud registration (I2P) is a fundamental task in robotic applications such as manipulation,grasping, and localization. Existing deep learning-based I2P methods seek to align image and point cloud features in a learned representation space to establish correspondences, and have achieved promising results. However, when the inlier ratio of the initial matching pairs is low, conventional Perspective-n-Points (PnP) methods may struggle to achieve accurate results. To address this limitation, we propose Angle-I2P, an outlier rejection network that leverages angle-consistent geometric constraints and hierarchical attention. First, we design a scale-invariant, crossmodality geometric constraint based on angular consistency. This explicit geometric constraint guides the model in distinguishing inliers from outliers. Furthermore, we propose a global-tolocal hierarchical attention mechanism that effectively filters out geometrically inconsistent matches under rigid transformation, thereby improving the Inlier Ratio (IR) and Registration Recall (RR). Experimental results demonstrate that our method achieves state-of-the-art performance on the 7Scenes, RGBD Scenes V2, and a self-collected dataset, with consistent improvements across all benchmarks.

The Hierarchical Kernel Transformer (HKT) is a multi-scale attention mechanism that processes sequences at L resolution levels via trainable causal downsampling, combining level-specific score matrices through learned convex weights. The total computational cost is bounded by 4/3 times that of standard attention, reaching 1.3125x for L = 3. Four theoretical results are established. (i) The hierarchical score matrix defines a positive semidefinite kernel under a sufficient condition on the symmetrised bilinear form (Proposition 3.1). (ii) The asymmetric score matrix decomposes uniquely into a symmetric part controlling reciprocal attention and an antisymmetric part controlling directional attention; HKT provides L independent such pairs across scales, one per resolution level (Propositions 3.5-3.6). (iii) The approximation error decomposes into three interpretable components with an explicit non-Gaussian correction and a geometric decay bound in L (Theorem 4.3, Proposition 4.4). (iv) HKT strictly subsumes single-head standard attention and causal convolution (Proposition 3.4). Experiments over 3 random seeds show consistent gains over retrained standard attention baselines: +4.77pp on synthetic ListOps (55.10+-0.29% vs 50.33+-0.12%, T = 512), +1.44pp on sequential CIFAR-10 (35.45+-0.09% vs 34.01+-0.19%, T = 1,024), and +7.47pp on IMDB character-level sentiment (70.19+-0.57% vs 62.72+-0.40%, T = 1,024), all at 1.31x overhead.

Long-context language modeling is commonly framed as a scalability challenge of token-level attention, yet local-to-global information structuring remains largely implicit in existing approaches. Drawing on cognitive theories of discourse comprehension, we propose HiCI (Hierarchical Construction--Integration), a hierarchical attention module that constructs segment-level representations, integrates them into a shared global context, and broadcasts both to condition segment-level attention. We validate HiCI through parameter-efficient adaptation of LLaMA-2 with only <5.5% additional parameters, extending context from 4K to 100K tokens (7B) and 64K tokens (13B). Across language modeling, retrieval, and instruction-following benchmarks, HiCI yields consistent improvements over strong baselines, including matching proprietary models on topic retrieval and surpassing GPT-3.5-Turbo-16K on code comprehension. These results demonstrate the effectiveness of explicit hierarchical structuring as an inductive bias for long-context modeling.

As long-context language modeling becomes increasingly important, the cost of maintaining and attending to large Key/Value (KV) caches grows rapidly, becoming a major bottleneck in both training and inference. While prior works such as Multi-Query Attention (MQA) and Multi-Latent Attention (MLA) reduce memory by sharing or compressing KV features, they often trade off representation quality or incur runtime overhead. We propose Memory-Keyed Attention (MKA), a hierarchical attention mechanism that integrates multi-level KV caches (local, session, and long-term) and learns to route attention across them dynamically. We further introduce Route-Fused MKA (FastMKA), a broadcast-routed variant that fuses memory sources before attention computation for improved efficiency. Experiments on different sequence lengths show that FastMKA achieves a favorable accuracy-efficiency trade-off: comparable perplexity to MLA while achieving up to 5x faster training throughput and 1.8x lower evaluation latency. These results highlight MKA as a practical and extensible framework for efficient long-context attention.