Linear Attention

KV-cache quantization and linear attention are two representative approaches to tackling the storage and computational costs of Transformers. KV-cache quantization compresses individual KV entries into discrete codes but retains all entries, whereas linear attention recurrently aggregates multiple historical KV contributions into a fixed-size continuous state but can introduce interference. This contrast raises the question of whether per-KV compression and multi-KV aggregation can be bridged within a single mechanism for efficient attention. We identify RAM-Net as such a bridge through soft assignments over a discrete address space. These assignments determine recurrent updates to the continuous slot state associated with each address. Under a restricted RAM-Net construction, we prove that soft address assignments extend hard quantized matching to a separable read-write overlap that locally approximates full-attention similarity and supports recurrent aggregation. These connections further enable Transformer-to-RAM-Net weight migration through a new path based on a soft-quantized intermediate construction. Across nine pretrained Transformer models from 0.3B to 7B parameters, RAM-Net recovers an average of 87.1% of the teachers' accuracy gains over random guessing across six commonsense and knowledge tasks using only a 500M-token budget per model.

Linear attention is a tractable model for understanding the mechanisms governing in-context learning in transformers. For linear regression tasks, recent asymptotic analyses have characterised its learning and generalisation behaviour. We extend this theory to nonlinear single-index targets, $y=f(x^\top w)+\varepsilon$. Our main result establishes a nonlinearity-noise equivalence: linear attention extracts only the linear Hermite component of $f$, while the remaining nonlinear structure contributes to the generalisation error as effective noise. This reduction allows results from the corresponding linear theory to be transferred to nonlinear tasks. We illustrate its implications for finite pretraining data and for the transition from task memorisation to task generalisation as task diversity increases. These results identify a limitation of the reduced linear-attention model and provide a tractable starting point for studying nonlinear in-context learning.

The architectural design of Large Language Models (LLMs) is shifting from traditional full-attention-only models to hybrid models, which combine different attention modules to improve long-context efficiency and performance in length extrapolation and context extension. To explain why hybrid models work and how to design them better, we propose Mechanics of Long-Context Hybrid Models. As Part 1.1 of this series, we begin with hybrids of full attention and either sliding-window attention (SWA) or gated variants of linear attention (LA), represented by GLA and GDN. We first observe a Seesaw Effect in Context Extension: LA hybrids benefit more from long-context continual pretraining, whereas SWA hybrids perform better under length extrapolation. We attribute this behavior to differences in the positional inductive biases induced by these attention mechanisms. We find that SWA hybrids suffer from a Short-Context Learning Trap, Short-Window Weariness, and Long-Window Laziness, and require extended windows to enhance performance in continual long-context pretraining. For LA hybrids, we summarize the Matthew Effect of Hybrid Position Extrapolation and propose Sliding-Window Linear Attention, achieving 16$\times$ training-free length extrapolation while maintaining 100% accuracy on NIAH-SK1 in 64k context length.

Hybrid architectures combining linear sequence models with softmax attention provide an effective balance between efficient long-context modeling and precise token retrieval. Existing designs such as Native Hybrid Attention (NHA) combine compressed long-term states with sliding-window attention, but their exact attention is restricted to a fixed local window. In this work, we introduce Prefix-State Hybrid Block Attention (PHBA), which replaces local sliding-window attention with top-k block-sparse retrieval and couples each retrieved block with a compact prefix state summarizing its preceding context. The prefix states are constructed by a gated linear recurrence at block boundaries and retrieved together with the corresponding token blocks, allowing the model to combine precise long-range evidence with compressed historical context within a unified layer. We further develop a hardware-aware Triton implementation that streams routed token blocks and prefix states without materializing large intermediate tensors. Experiments show that PHBA improves long-context and retrieval performance over strong linear and hybrid baselines while retaining efficient training and inference.

Linear attention enables efficient long-context autoregressive decoding by compressing history into recurrent states, but this compression can make selective access to sparse and distant information difficult. Existing chunk-based extensions increase memory capacity, yet learned chunk-mixing coefficients may remain fixed with respect to input content and therefore cannot adapt historical access to each query. We introduce \emph{Hybrid Linear Attention} (HLA), a query-dependent chunk-level attention mechanism for Gated DeltaNet (GDN). HLA represents each completed chunk as an exact affine state transition and computes content-dependent routing gates from compact, self-attentively pooled representatives. Each gate interpolates the corresponding historical transition with the identity map, controlling both the chunk's additive memory and its transformation of earlier states. Effective-support regularization further encourages concentrated routing for sparse inference. We evaluate HLA under both pretrained adaptation and from-scratch training. Across Qwen3.5 models from 0.8B to 9B, HLA consistently improves over native GDN and fixed chunk mixing, with gains of up to 5.57 percentage points on LongBench-V2 and 3.97 points on RULER. In a controlled from-scratch 1.3B setting trained for 100B tokens with a 4K context, HLA also improves RULER performance from 4K to 32K, with gains increasing from 0.83 points at 4K to 4.22 points at 32K. These results demonstrate that query-dependent composition of recurrent memory improves long-context modeling and remains effective beyond the training context while using compact per-chunk affine summaries. Project page: https://caesarhhh.github.io/hla/

Long-horizon video world models require persistent memory to preserve scene consistency over extended rollouts. Softmax attention retains the full generation history through a growing KV cache, whereas recurrent linear attention compresses history into fixed-size states with substantially lower memory cost. However, we identify severe long-range forgetting in Gated DeltaNet (GDN), where information from distant but relevant scenes is progressively attenuated by subsequent state updates. To address this limitation, we propose HLA-WM, a training-free hybrid linear-attention framework that combines coarse-grained geometry-guided retrieval with fine-grained recurrent linear-state computation. HLA-WM exploits the affine structure of GDN to cache compact chunk-wise transition summaries, retrieve scene-relevant historical chunks using camera geometry, and recompose them into query-specific recurrent states. On the $60$-second SANA-WM-Bench, HLA-WM improves all six aggregate revisit-consistency and camera-control metrics of the base autoregressive generator without additional training, including a $0.74$ dB PSNR gain and a $28.5\%$ reduction in rotation error. The improvements persist after downstream refinement and generalize to MBench-A, where HLA-WM consistently improves all three revisit-consistency metrics across all four subsets and all evaluated inference modes over $547$ samples. At a $60$-second context, HLA-WM reduces historical-state memory by $12\times$ relative to full KV caching while incurring at most a $1.6\%$ reduction in inference throughput. These results demonstrate that selectively addressable recurrent memory can improve long-range scene recall while preserving the efficiency advantages of GDN. Project page: https://caesarhhh.github.io/hla-wm/

Designing expressive sequence layers with efficient inference remains a central challenge in modern machine learning. Standard softmax attention achieves excellent sequence modeling performance through rich nonlinear token interactions, but it requires a key-value cache that grows linearly with sequence length, limiting its scalability. Linear attention enables efficient recurrent computation with a constant memory footprint, yet its reduced expressivity often yields inferior modeling performance. We introduce Switching Linear Attention (SwiLA), a novel sequence layer that bridges this gap by enhancing representational capacity while retaining the fixed-size recurrent state of linear attention. We derive the SwiLA recurrence from the test-time regression framework, casting the state update rule as online expectation-maximization in a mixture of linear regressions model. At test time, each output dimension dynamically selects among multiple linear attention components based on the input. Across associative recall, in-context language learning, and language modeling benchmarks, SwiLA shows strong performance and narrows the gap to softmax attention, even surpassing it in several settings.

Long-context sequence models face a fundamental tradeoff: softmax attention uses flexible token-level interactions at quadratic cost, whereas linear attention obtains linear-time training and constant-time decoding by compressing history into a fixed-size state. In this work, we ask whether we can connect these regimes through a tunable notion of structure. To this end, we introduce Structured Matrix Attention (SMat-Attention) via a family of causal masks with structured long-range routing whose row supports have VC-dimension $d$. In our construction, $d=1$ recovers the standard causal mask, and increasing $d$ permits richer subset-routing patterns. We give chunkwise forward and backward algorithms to enable hardware-efficiency. For sequences of length $T$, the hard-routing construction takes $O(T^{2-3/d}+T)$ work, despite the mask being dense, for our prescribed family. In fixed-horizon streaming, decoding after the distant prefix takes constant time per token using $O(T^{1-1/d})$ cached states. SMat-Attention therefore makes VC-dimension an explicit knob governing access-pattern complexity, prefill cost, and decoding memory. Empirically, subset-routing and rule-assisted multi-key retrieval experiments illustrate the masks' routing expressiveness. Extensions to Mamba-2 and Gated DeltaNet using learned routing with top-$k$ query reads retain subquadratic prefill, improve recall accuracy over the backbones in several settings, and achieve comparable small-scale language-modeling performance.

Gated Linear Attention (GLA) Transformers advance linear recurrent models through data-dependent gating, but face a core limitation: the fixed-capacity memory matrices across all heads operate at a single temporal resolution, where each token is processed individually, forcing them to simultaneously encode local syntactic patterns and long-range semantic structure, creating a representational bottleneck that gating alone is insufficient to resolve. We introduce Multi-Scale Gated Linear Attention (MS-GLA), which addresses this by distributing attention heads across multiple temporal resolutions. Coarser resolutions pool longer token spans naturally specializing toward long-range dependencies, while finer head groups retain sensitivity to local syntactic structure. A learnable, input-dependent fusion layer dynamically recombines head group outputs at each timestep, expanding effective memory capacity without increasing per-head state size. This multi-resolution decomposition draws on principles from Multi-Scale State-Space Models (MS-SSM), adapting them to the gated linear attention setting. We evaluate MS-GLA on language modeling, recall-intensive tasks, and long-context generalization. Across all settings, MS-GLA consistently achieves higher accuracy and lower perplexity than GLA at matched parameter counts, with up to 18.9% improvement on recall-intensive tasks and 9.5% lower average perplexity on language modeling benchmarks, validating multi-temporal resolution decomposition as a principled and effective extension of Gated Linear Attention.

Hybrid-attention models replace most softmax attention layers with linear attention, reducing KV-cache growth and enabling larger decode batches where recurrent-state access becomes a major bottleneck. ReplaySSM amortizes state updates by buffering keys and values, but each new query still requires a full-state read even though the state remains unchanged between state updates. We observe that low-rank state-weighted query approximation accurately preserves state-read outputs. Although future queries are unknown, the basis vectors used to approximate them can be fixed offline. Based on this observation, we introduce SketchSSM, which preserves full-state updates while approximating reads. At each state update, SketchSSM reads the full state once to precompute outputs for these basis vectors, storing them in a compact sketch. Each subsequent decode step combines the sketch vectors with query-dependent coefficients to reconstruct the output without a full-state read. Across four Mamba-2-, GDN-, and KDA-based models, SketchSSM at a mean sketch rank of 8 reduces state-access traffic by approximately 10$\times$ while matching the average accuracy of the FP32 full-state baseline across four decode benchmarks, and preserves recall on four RULER retrieval tasks. At this rank on one NVIDIA B300, linear-attention kernel speedups over the Standard vLLM baseline reach 7.30$\times$, 5.02$\times$, and 5.24$\times$ for Mamba-2, GDN, and KDA, respectively, with up to 2.77$\times$ higher decode throughput on Nemotron 3 Super.

Self-attention-based vision backbones perform well on dense prediction, but the quadratic computational cost of global softmax attention limits their efficiency as image resolution increases. We introduce Gated Token Recurrence (GTR), a softmax-free recurrent vision backbone that combines gated linear attention, alternating spatial scan directions, and spatially enhanced SwiGLU blocks. GTR is distilled from a detection-specialized DINOv3 teacher using only final-layer patch-token alignment through a linear projection and squared $\ell_2$ loss, without masked-token prediction or intermediate-layer supervision. With Objects365 detector pre-training, GTR-L achieves 58.9 box AP on COCO \texttt{val2017} with 1.908,ms median batch-one latency under compiled FP16 execution on an RTX4090. The same backbone also transfers to instance segmentation, pose estimation, oriented detection, semantic segmentation, and monocular depth estimation. In an isolated kernel benchmark, our specialized chunkwise CUDA operator is $4.0\times$ faster than FLA v0.5.0 at 1.6K tokens on RTX4090. TensorRT deployment on DRIVE AGX Thor achieves 2.282--8.769,ms median batch-one latency across the evaluated models. These results show that recurrent token mixing can provide an efficient alternative to global softmax attention for high-resolution dense prediction and edge deployment. Project page: https://intellindust-ai-lab.github.io/projects/GTR/

Vision Transformers devote most of their parameters to MLPs for channel mixing, but still rely on quadratic multi-head self-attention for token interactions. While linear attention fixes the complexity problem, bringing it down to O(N), it is usually just paired with the same fixed-activation MLP as before. Kolmogorov-Arnold Networks take a different approach, placing learnable univariate functions on the edges instead. However, existing vision KANs either retain standard attention or remove attention entirely, so the two ideas have not been effectively combined. We introduce LKAT (Linear Kolmogorov-Arnold Transformer) to close this gap: an isotropic ViT-style encoder that couples chunk-wise Gated Linear Attention with a two-layer KAN feed-forward block, backed by an I/O-aware fused RBF-KAN kernel to make radial-basis grid functions efficient in practice. Under a shared DeiT-style training recipe, LKAT-B outperforms ViT-B/16, ViT-5-B, and Mixer-B/16 on ImageNet-100, while Tiny, Small, and Base variants scale consistently on CIFAR-10/100. ImageNet-100 pretraining also transfers effectively to CIFAR fine-tuning, suggesting that gated linear attention and KAN-based radial basis functions provide complementary inductive biases for mid-scale visual representation learning. Code: https://github.com/mehizelali/linear-kan-transformer

Video diffusion models repeatedly process long spatiotemporal token sequences during denoising, making attention a major computational bottleneck. Linear attention offers an appealing alternative and has been widely adopted in recent large language models, but directly applying it to video models often fails to preserve the fine-grained interactions required for high-quality generation. We present Video DeltaNet (VDN), which combines local Softmax attention with bidirectional linear memory for long-range video context. Its linear branch introduces Video Delta Attention (VDA), which updates memory once per frame by jointly incorporating its spatial tokens. Separate output projections and learnable gates calibrate the two branches, while a staged teacher-alignment recipe progressively introduces the new pathway into pretrained models. We instantiate VDN on MiniMax H3, applying the hybrid to video-to-video interactions while retaining Softmax for interactions involving text or audio. With eight-step distillation and an optimized SGLang serving stack, VDN-H3 completes DiT denoising for a 14.3-second, 768p video in 6.70 seconds on eight NVIDIA B200 GPUs, corresponding to a 14.5x speedup over the 50-step dense H3 baseline on the same GPU count. GitHub code available at: https://github.com/OpenVDN/vdn-minimax-h3. Weights available at: https://huggingface.co/OpenVDN/vdn-minimax-h3

Generative recommendation (GR) relies on large-beam decoding to generate hundreds of candidate items, creating a new scaling challenge for recurrent linear attention. Existing linear attention serving systems either materialize a full recurrent state for every beam or repeatedly replay shared history, incurring substantial memory and traffic overhead. To address this, we present OneLA, a linear-attention decoding framework that exploits the shared prompt and short divergent suffixes of GR workloads. Specifically, OneLA represents all beam states using a single shared prompt-derived state and compact, append-only records of their divergent transitions. Using this representation, OneLA computes only the state information required at each decoding step, without reconstructing a full recurrent state for every beam. Furthermore, OneLA uses a lightweight ancestry index to track the transition records that make up each beam's history, allowing beams to be updated without moving or copying existing records. A fused GPU kernel further reuses the shared state across beams. Our analysis shows that OneLA achieves 1.54-2.46x end-to-end decode speedups while substantially reducing recurrent-state memory use and data movement.

Linear attention and state-space models provide linear-time sequence modeling, but their recurrent memory remains a second-order tensor (a matrix), limiting the order of interactions that can be represented in the state. We introduce the RunningTensor, which generalizes this memory to an order-$o$ tensor, updated by a rank-1 outer product and read by contracting against $o-1$ vector queries. Order $2$ recovers linear attention; we study order $3$ as a proof of concept, retaining both recurrent and parallel forms while remaining linear in sequence length $T$ and improving working memory capacity from $\mathcal{O}(W^2)$ to $\mathcal{O}(W^o)$. On synthetic multi-query associative recall, RunningTensor outperforms linear-attention and SSM baselines. After pretraining, it also improves performance on language-understanding and non-synthetic retrieval tasks, suggesting that higher-order recurrent state can provide useful additional memory capacity beyond matrix-valued state.

Linear attention enables efficient long-context inference by compressing token history into a fixed-size recurrent memory. This compression makes each update a trade-off between incorporating new information and preserving useful associations. Models such as DeltaNet, Gated DeltaNet, and KDA predict write strength from the current token representation, without explicitly tracking uncertainty in the stored memory. Yet this uncertainty matters: a new observation should have greater influence when the existing association is uncertain and less when it is already well supported. We introduce Kalman Delta Networks (KDNs), a family of linear-attention models that explicitly track memory uncertainty to guide each update. By formulating associative memory as a linear-Gaussian state-space model, KDNs propagate both the memory estimate and its uncertainty, using the Kalman gain to balance accumulated evidence against the reliability of new observations. This formulation also recovers standard delta-rule updates by replacing tracked covariance with a token-predicted isotropic surrogate. To support hardware-efficient training and inference, we derive Diagonal KDN and Isotropic KDN, which retain one uncertainty value per key channel and per head, respectively. Their uncertainty updates admit associative scans with logarithmic parallel depth, requiring only $O(d_k)$ and $O(1)$ auxiliary state per head. Across controlled pretraining at 750M and 1.3B parameters, both variants consistently improve perplexity and mean downstream accuracy over the evaluated state-of-the-art linear-attention baselines.

Diffusion Transformers (DiTs) achieve strong video generation quality, but their dense spatiotemporal self-attention scales quadratically with sequence length and quickly becomes the dominant inference bottleneck. Sparse low-rank hybrids alleviate this cost by combining a local sparse branch with a global compressed branch. In video DiTs equipped with 3D Rotary Position Embeddings (RoPE), the global branch faces a structural compatibility issue: when RoPE is applied before a nonlinear feature map, the rotation and nonlinearity generally do not commute, making it difficult to keep a query-independent linear summary while preserving relative rotary geometry. Existing work often sidesteps this issue by replacing genuine cross-token global aggregation with coordinate-conditioned surrogates or learnable absolute positional modules. These compromises can be effective, but they approximate relative decay from absolute coordinates and introduce extra positional parameters. We propose \textbf{RoLA}, a rotary-positioned low-rank linear-attention branch that keeps genuine cross-token aggregation while remaining compatible with a reusable linear summary. The design applies RoPE \emph{outside} the nonlinear low-rank feature map and reuses a truncated subset of the pre-trained rotary schedule matched to the low-rank bottleneck. This yields a linear-time low-rank global branch with relative positional behavior by design and no additional positional parameters; the full sparse--low-rank module still includes the fixed-sparsity sparse branch. Experiments on open-source video DiTs show that the resulting method remains competitive in generation quality at 90% sparsity while achieving 2.63$\times$ end-to-end inference speedup on Wan2.1-14B (720p, 81 frames, measured on an NVIDIA H100 GPU).

Hybrid architectures combining Full Attention (FA) and Linear Attention (LA) are increasingly prominent, yet their allocation remains heuristic. We seek an evidence-grounded basis in head-level functional organization learned by RoPE-based Transformers. Behavioral probes do not yield a complete taxonomy, so we propose two intervention metrics: RoPE Frequency Importance Score (RFIS), measuring how each frequency affects a head's attention distribution, and RoPE Positional Dependence (RPD), isolating dependence on rotary positional modulation. On Qwen3-series models and Llama3.1, RFIS suggests and RPD verifies a complete taxonomy of retrieval and positional heads separated by a salient mid-low-frequency band. Controlled Transformers show that this boundary follows the training-length positional scale; we term it the Global Positional Band (GPBand). The analysis suggests a potential cause of zero-shot length-extrapolation failure and yields two principles: positional modeling should operate only locally, with global access through position-independent retrieval; and both functions should be assigned at head granularity with layer-specific allocation. We instantiate them in Head-wise Hybrid Architecture (HwH), using NoPE FA for global retrieval and LA for local positional modeling. With an FA-to-LA ratio below 1:3, HwH retains strong language modeling and commonsense reasoning while improving retrieval and substantially strengthening zero-shot long-context extrapolation over Transformer, LA, and a layer-wise hybrid baseline. Ablations validate both principles and component roles, highlighting principled hybrid architecture design as a promising route toward future foundation models.

Real-world time series often exhibit irregular sampling and extended temporal horizons, requiring models to capture continuous-time dynamics across arbitrary intervals without prohibitive scaling costs. Discrete-time methods collapse variable time intervals into static positional steps; solver-dependent continuous-time models preserve temporal structure but rely on sequential integration, precluding parallelization; and solver-free approximations avoid this cost yet none couples observed time intervals with input-driven state modulation. We propose Liquid Gated Attention (LGA), a solver-free parallel temporal operator. By parameterizing an input-driven gating mechanism with observed time intervals, LGA introduces a continuous-time inductive bias and formulates hidden state evolution as a fast-weight associative memory, enabling parallel computation across the temporal dimension. Using matrix associativity in non-causal encoding and a prefix scan in causal encoding, LGA attains linear temporal complexity in sequence length in both modes. A sequence-level normalization bounds cumulative temporal decay for stable long-horizon optimization. Building on LGA, we instantiate LFormer, a modular backbone for continuous-time representation learning. Across six tasks and sixteen datasets spanning up to 17,984 steps, LFormer demonstrates long-range dependency modeling, fine-grained state tracking, and trajectory reconstruction from sparse and noisy observations, while delivering competitive performance against state-of-the-art discrete-time and continuous-time baselines with linear scaling efficiency.

Due to the nature of quadratic attention, Large Language Models (LLMs) consume a lot of memory and energy: every new token costs more than the previous one, and its keys and values must be stored in memory indefinitely, which is unsustainable. Two main lines of work address this: compressing the KV cache, e.g., by evicting or quantizing keys and values, and retrofitting LLMs to use Linear Attention, which replaces the KV cache with a fixed-size state. Retrofitting has attracted a lot of attention, given its promise to solve the quadratic scaling problem with state-of-the-art performance at low cost. However, it has not been properly compared to the simplest form of KV-cache eviction: Sliding Window Attention (SWA) with attention sinks. In this work, we show that SWA with sinks performs as well or better than most retrofitted Linear Attention models across multiple LLMs and downstream tasks, with the largest gains on long-context and generative tasks. On long-context reasoning tasks (Needle-in-a-Haystack and BABILong), SWA achieves massively higher performance (2 to 10 times higher than linear attention). SWA requires no additional training, is extremely fast, and requires little memory, making it an extremely cheap and reliable solution. When the training budget is limited, switching to SWA is a much more effective way to reduce inference memory cost than retrofitting linear attention. Linear attention models have shown promise, but they require training from scratch or extensive retrofitting to reap their architectural benefits and come close to SWA.

We present the first systematic study of Massive activations (MAs) in layer-interleaved HLA LLMs and uncover two architecture-aligned morphologies: MAs consistently spike immediately before full attention layers, forming pre-attention spikes (PAS), and can persist through intervening linear attention layers, giving rise to inter-spike plateaus (ISP). As full attention becomes denser, successive PAS become increasingly connected through ISP, ultimately recovering the stable MA morphology of full attention LLMs. We establish the recurrence of this organization across five linear attention architectures, six hybridization configurations, five data domains, and representative open-source hybrid models spanning 1.2B to 397B total parameters. Controlled pretraining of GDN-based hybrids at scales up to 1.3B shows that both morphologies emerge early and respond asymmetrically to output gating: full attention output gating strongly attenuates their absolute magnitudes without eliminating their layerwise organization, whereas removing GDN gates yields comparatively modest amplification. Mechanistically, our systematic-outlier analysis supports a shared lifecycle account governed by the timing of MA cancellation. PAS follows a localized write-sink-cancel process, while the extended persistence of ISP is consistent with delayed cancellation. At the full attention limit, this account recovers the stable MA morphology characteristic of full attention LLMs. Our code is available at https://github.com/StartluxLabs/Massive-Activations-HLA.

State Space Models (SSMs), as a mainstream research direction of linear Transformers, aim to achieve higher efficiency than standard Transformers in long-context modeling. However, existing SSMs suffer from limited input adaptivity and constrained memory capacity, leading to information loss when modeling ultra-long sequences. To address these limitations, we propose MixFormer, a novel linear Transformer that integrates a Mixture-of-Memory-Experts (MoE) mechanism. Specifically, the model maintains differentiated memory states through multiple collaborating memory experts and employs a novel Time-Aware Linear Attention (TALA) mechanism, which leverages learnable exponential decay functions and positional biases to dynamically update memory. This design enables the model to selectively reinforce important historical information while effectively mitigating memory dilution, substantially improving long-range dependency modeling. Experiments on long-sequence text and image generation tasks demonstrate that MixFormer not only achieves significant performance gains but also provides a more sustainable computational backbone for the next generation of web infrastructure.

We present a linearized form of 2-simplicial attention by rewriting the trilinear score as an inner product between a composite query and a key, so that the sum over one token axis takes the same form as ordinary softmax attention. We then approximate this sum with positive random features and store the entire past in a fixed-size state, while the second axis stays explicit over a short window of recent tokens. This enables us to achieve linear cost in sequence length combined with a global reach that windowed 2-simplicial attention lacks. We implement it with custom Triton kernels and combine it with Kimi Delta Attention to build a model with no softmax attention at all. Under matched compute, this model achieves the highest mean downstream accuracy among the compared architectures, and at 16k context it improves mean accuracy over a KDA hybrid while lowering LAMBADA perplexity from 715.6 to 602.6.

Vision transformers face significant computational overheads in high-resolution dense prediction due to the quadratic complexity of self-attention. Linear attention offers efficiency but sacrifices local context modeling. We propose \textbf{HSMLA (Hierarchical Softmax Multi-scale Linear Attention)}, which combines ReLU-based linear attention for global context, selective softmax refinement for critical local features, and multi-scale token representations via depthwise convolutions. HSMLA achieves superior accuracy-efficiency trade-offs: up to $4.2\times$ inference-time speedup across dense prediction tasks, $87.3$ Dice with $3.2\times$ speedup on CT organ segmentation, and $94.2$ AUC with $4.1\times$ speedup on pathology WSI.

Diffusion language models (dLLMs) offer a promising alternative to autoregressive models by accelerating inference through parallel decoding. Recent dLLMs commonly use blockwise semi-autoregressive decoding, generating blocks autoregressively while denoising tokens within each active block in parallel. However, despite KV caching, each denoising step still attends to all previous blocks, repeatedly incurring prefix-attention cost. Motivated by this bottleneck, we ask whether dLLM inference can be further accelerated by linearizing attention over previous blocks. We introduce block-hybrid attention, which retains exact softmax attention within the active denoising block while applying linear attention over previous blocks. We show that this hybrid attention can be retrofitted into a pretrained dLLM with minimal post-training: LLaDA-Hybrid replaces 6 of the 20 attention layers in LLaDA~2.1, a 16B open-source dLLM, largely following LoLCAT (Zhang et al, 2024). The conversion takes only approximately 60 hours while preserving benchmark performance: 72.0% vs. 75.6% on HumanEval, 63.0% vs. 57.7% on MBPP+, and 86.7% vs. 88.3% on CMATH. With a Triton implementation, LLaDA-Hybrid achieves up to $1.7\times$ higher decoding throughput and supports more concurrent requests before exhausting memory, showing that pretrained dLLMs can be efficiently linearized for faster inference. Our code is available at: https://github.com/Diuven/LLaDA-Hybrid.

Laser line-scanning microscopy enables fast volumetric imaging but produces anisotropic lateral resolution. Orthogonal line scans provide complementary directional information that can recover near-isotropic resolution, yet existing deep-learning methods require a separate model for each optical configuration. We present a unified, resolution-conditioned fusion framework based on Rank Enhanced Linear Attention (RELA). Feature-wise Linear Modulation (FiLM) conditions the network continuously on the resolving-power ratio, enabling one model to adapt across slit widths. We further introduce Adaptive RELA, which replaces fixed-kernel rank enhancement with ratio-conditioned multi-scale depthwise convolutions and uses a learnable attention temperature to adjust selectivity with degradation severity. Training data spanning multiple slit configurations are generated using a physics-grounded separable point-spread-function model verified against measured optical data at 48.3 dB accuracy. The resulting model achieves 34-40 dB PSNR across configurations, whereas unconditioned multi-slit training collapses to 24.3 dB and per-slit specialists lose 4-9 dB outside their training setting. It also generalizes smoothly to unseen intermediate configurations without interpolation artifacts. Ablations show that FiLM resolves configuration ambiguity, global linear attention captures long-range directional correspondences, and adaptive temperature yields an additional 2 dB in the challenging near-isotropic regime, where complementary signals are weak.

Hybrid-attention large language models combine full attention with recurrent linear attention to reduce long-context inference costs, yet their autoregressive decoding remains memory-bound. Tree speculative decoding offers an attractive acceleration path, but existing tree-speculation systems are designed around the key--value caches of full-attention models. On hybrid models, they traverse recurrent layers branch by branch and materialize a full state for every proposal node, causing verification latency and transient memory to scale poorly with tree and batch sizes. We present Bole, a kernel--runtime co-design that enables efficient tree speculation for hybrid-attention LLMs. Bole transforms the linear-attention recurrence into a tree-structured closed form and realizes it with a resource-efficient GPU kernel, verifying all proposal nodes in parallel and accelerating linear-attention tree verification by 3.4--7.7$\times$. It losslessly encodes speculative state updates as token-level factors and reconstructs only the state selected after sampling, reducing transient state memory by 82--99$\times$ and freeing GPU capacity for KV caches. Its integration into SGLang, a widely deployed production LLM serving engine, couples efficient state management with a batch-wide verification budget calibrated to the complete hybrid forward. Across four models, two GPU platforms, and diverse datasets, Bole delivers up to $4.72\times$ the offline decode throughput of autoregressive decoding and up to $2.03\times$ that of the strongest tree-speculative baseline. Under online agent workloads, it reduces TTFT and TPOT by up to $67.6$ and $49.9$, respectively, over the strongest tree-speculative baseline.

Vision-Language Models (VLMs) face a critical computational bottleneck when processing high-resolution imagery due to the $O(N^2)$ memory complexity of Softmax Multi-Head Attention (MHA). While substituting MHA with independent Multi-Layer Perceptrons (MLPs) achieves $O(N)$ scaling, it strips the architecture of spatial sequence routing, severely degrading global scene understanding and object permanence. In this paper, we propose the Linear Multi-Timescale Retention (LIA-MTR) module, a memory-efficient cross-modal bridge. By integrating an ELU-based positive feature mapping with adaptive write-gating and log-linearly distributed recurrent decays, LIA-MTR mathematically compresses continuous visual sequences into bounded memory states. Theoretical analysis proves the architecture operates with strict $O(N)$ sequence-interaction complexity. Empirically, synthetic retrieval evaluations demonstrate that LIA-MTR flawlessly routes context across 16,000 tokens, eliminating the "Lost in the Middle" degradation typical of naive linear attention. Hardware benchmarking reveals infinite-context scaling capabilities, natively processing 262,144 visual patches within an 11.2 GB VRAM footprint, whereas standard MHA suffers out-of-memory failure at 16,384 patches. Furthermore, following instruction tuning on 665K conversational samples, LIA-MTR significantly outperforms an industry-standard MLP baseline on the MME benchmark (71.00% vs. 68.11%), driven by a 10% absolute improvement in object permanence and superior global semantic extraction. This work establishes a mathematically rigorous, computationally flat foundation for infinite-context Vision-Language integration.

We study information bottlenecks in modern deep-learning architectures -- RNNs, softmax transformers, linear-attention transformers and state-space models -- through the lens of the indexing primitive. In this primitive, the input consists of $n$ bits and one integer $i$ from $1$ to $n$ called the index, and the output equals the value of the $i$-th bit. We introduce causal complexity for masked architectures. We show that architectures with low causal complexity cannot solve the indexing primitive in any constant number of layers when the index appears at the end of the input. In particular, this limitation applies to low-parameter RNNs, SSMs and masked linear-attention transformers. In contrast, small softmax transformers can solve it in one layer, while non-masked linear-attention transformers can solve it in 2, which separates them from their masked counterparts. In turn, when the index appears at the beginning, we show that small RNNs are capable of solving this task in 1 layer, while all the other architectures require 2. All our impossibility results are unconditional and apply even to models that employ infinite-precision real arithmetic. Moreover, experiments for up to $n=64$ qualitatively align with our theory: configurations with low-parameter theoretical solutions learn the indexing task easily, while configurations that do not admit such theoretical solutions struggle to learn as the sequence length grows.

We introduce SANA-Video 2.0, a hybrid video diffusion transformer instantiated at 5B and 14B scales under a unified architecture. Designed to generate high-quality video up to 720p on a single GPU, SANA-Video 2.0 matches full-softmax video DiTs in quality while retaining the favorable long-sequence scaling of linear attention. To avoid quadratic attention throughout, Hybrid Linear-Softmax Attention combines gated linear attention for O(N)-dominated mixing with periodic gated-softmax anchors at a 3:1 ratio, restoring the full-rank token interactions that pure linear attention lacks. To propagate these refreshed representations across depth, Block Attention Residuals (AttnRes) route completed block summaries into later linear layers, enabling anchor-feature reuse and boosting deep-layer effective rank by ~12%. Through from-scratch training, SANA-Video 2.0 learns the complete hybrid directly rather than linearizing pretrained models, with reduced-resolution proxy studies establishing 25% softmax as the optimal quality-efficiency trade-off. With 40-step sampling, SANA-Video 2.0 achieves a VBench score of 84.30 in 13.2s at 480p on a single H100, remaining competitive with far larger softmax video DiTs at a fraction of the latency. Its compiled DiT forward pass is 3.2x faster than a matched full-softmax baseline at 720p/60s, a gap that expands with video duration. Furthermore, full-stack Sol-Engine optimization (kernel fusion, caching, and sparse attention) accelerates this hardware-friendly backbone by a further 3.58x, bringing the 5B pipeline to 13.06s at 720p/5s and making it 120x faster than Wan 2.2-A14B on one H100. Overall, our hybrid design recovers softmax-level expressiveness at substantially reduced cost, unlocking scalable long, high resolution video generation.