每日文章

Recently, end-to-end OCR models, exemplified by DeepSeek OCR, have once again thrust OCR into the spotlight. A widely held view is that employing a large language model (LLM) as the decoder allows the model to leverage the prior distribution of language, leading to improved OCR performance. However, the downside is equally evident: as the output sequence lengthens, the accumulated KV cache drives up memory consumption and progressively slows down generation. This stands in stark contrast to humans, who exhibit no such decline in efficiency during long-horizon copying tasks. In this technical report, we propose Unlimited OCR, a model designed to emulate human parsing working memory. Taking DeepSeek OCR as the baseline, we replace all attention layers in the decoder with our proposed Reference Sliding Window Attention (R-SWA), which reduces attention computation costs while maintaining a constant KV cache throughout the entire decoding process. By combining the high compression rate of DeepSeek OCR's encoder with our constant KV cache design, Unlimited OCR can transcribe dozens of pages of documents in a single forward pass under a standard maximum length of 32K. More importantly, R-SWA is a general-purpose parsing attention mechanism - beyond OCR, it is equally applicable to tasks such as ASR, translation, etc. Codes and model weights are publicly available at http://github.com/baidu/Unlimited-OCR.

Efficient sampling of the Boltzmann distribution in frustrated spin glasses is central to statistical mechanics and combinatorial optimization. Despite advances in machine-learning-based approaches, two issues persist: limited understanding of why variational models fail to benefit from increased scale, unlike the monotonic scaling law of large language models; and high computational cost on large systems that negates advantages over classical sampling methods. Here, we develop a physics-inspired transformer with interpretable sparse attention and spin-tailored positional embeddings to address these challenges. By further leveraging FlashAttention for parallel ancestral sampling, it achieves up to two orders of magnitude speedup over vanilla variational autoregressive networks, enabling neural-network simulations of spin-glass systems to unprecedented sizes on a single GPU. It can resolve full probability distributions, free energies, and overlap statistics across temperatures, for Sherrington-Kirkpatrick and 2D or 3D Edwards-Anderson models, where existing machine-learning methods encounter limitations at certain temperatures. This framework thus establishes a scalable paradigm for frustrated spin-glass systems.

Latent diffusion models achieve strong generative performance by operating in a compressed latent space produced by a variational autoencoder (VAE). However, it remains unclear whether all latent channels contribute equally to the diffusion process, or whether significant redundancy exists. We introduce PeLAP-A (Adaptive Latent Pruning for Diffusion), a lightweight framework that augments a standard latent diffusion pipeline with a learnable channel-wise importance predictor. A two-layer MLP operating on globally pooled latent features produces a soft mask that suppresses unimportant latent channels before they enter the denoising UNet. The entire system is trained jointly on CIFAR-10 under a combined diffusion, reconstruction, and sparsity loss. Experiments reveal a striking result: under aggressive sparsity regularization (lambda = 0.01), the importance predictor drives all latent channels to near-zero yet the denoising UNet achieves lower diffusion loss (0.0236 vs. 0.0240) and lower VAE reconstruction MSE (22.59 vs. 24.67) compared to the unpruned baseline. We term this the sparsity collapse phenomenon and provide an analysis of why it occurs and what it reveals about the information requirements of latent diffusion models. These findings constitute an exploratory study of sparsity dynamics in latent diffusion training, and demonstrate that denoising UNets can remain remarkably robust to latent channel suppression even under aggressive regularization. Code is available at: https://github.com/kissasium/PeLAP-A.git.

Multimodal agents repeatedly re-examine the same video frames, UI screenshots, and rendered artifacts as their context window slides and reasoning iterates, yet every look-back re-encodes from scratch, because prefix caches serve reuse only at a fixed leading position. We show this recompute is avoidable, and identify exactly what naive KV reuse loses: the cross-chunk conditioning a chunk absorbs from its neighbours. This loss is asymmetric. The direct readout of a cached chunk is recovered exactly and for free by the standard state-merge. What remains is a diffuse, low-rank residue concentrated in deep layers, invisible to single-hop retrieval but precisely what multi-hop reasoning binds on. Blind reuse therefore leaves single-hop recall intact while halving multi-hop accuracy; this is the failure mode prior position-independent caches, designed for single-context or single-image reuse, do not address. We repair it with a small, training-free low-rank conditioning patch stored alongside each position-free chunk. Reuse reduces to one operator across MLA, GQA, and MHA: exact RoPE re-rotation to any target position, plus the patch that restores cross-chunk binding. This makes three window operations cheap: reorder (one patch serves every ordering of a cached set), sliding-window survival (surviving chunks relocate via rotation only, zero re-encode), and recall (an evicted chunk is rehydrated by its patch, never re-encoded). A rank-m patch recovers full task accuracy on cross-chunk-binding benchmarks, MM-NIAH across two attention families and two-page doc-QA, at a fraction of the KV footprint, and reconstructs re-prefill KV to within bf16 rounding in a production SGLang kernel across six backbones. The conditioning signal is strongest in redundant vision and video streams, making our solution most impactful where multimodal agents spend their recompute budget.

Remote photoplethysmography (rPPG) estimates the blood volume pulse (BVP) signal from facial videos, enabling contact-free health monitoring. Conventional clip-wise approaches, which use video clips as input, require capturing over one hundred frames before inference, thus introducing several seconds of delay and hindering real-time use. Meanwhile, frame-wise approaches struggle to capture long-range temporal and periodic features of physiological rhythms, and therefore lead to reduced estimation accuracy. To overcome these issues, we propose StreamPPG, a unified architecture that enables low-latency frame-wise physiological signal estimation while achieving competitive accuracy compared with clip-wise approaches. StreamPPG is trained under a consistent privileged learning (CPL) strategy, which leverages ground-truth rPPG signals as privileged information to enhance the model's representation capability. Extensive experiments demonstrate that StreamPPG achieves state-of-the-art accuracy across multiple datasets while maintaining real-time throughput on edge devices.

Long-running LLM agents keep valuable state resident on GPUs: KV caches, request schedulers, communication state, and sometimes online adapters. Losing this state after a GPU or communicator failure can discard minutes to hours of work, yet existing recovery mechanisms either restart the whole serving stack or require application-specific checkpoint logic inside every attention and runtime component. This paper argues that fault tolerance for such workloads needs a GPU-resident execution context: checkpoint hooks must run at device synchronization points, observe binary kernels that frameworks and libraries actually execute, and recover without putting the host CPU on the critical path. We present Concordia, a runtime that uses a device-resident persistent kernel as the substrate for fault-tolerant LLM inference. Concordia interposes on GPU module loading and supports PTX- and SASS-level instrumentation, allowing checkpoint and pause hooks to be inserted below framework code and library boundaries. For each registered LLM state region, Concordia JIT-compiles a specialized delta-checkpoint handler -- for example, a KV-block scanner, adapter-page scanner, or recovery applier -- and hot-swaps it into the persistent kernel's operator table. The persistent kernel consumes a lock-free ring buffer of compute, checkpoint, append-log, and recovery tasks, so the same always-on executor triggers dirty-page detection, stages deltas, and appends committed records to a CPU-visible log in CXL memory or host DRAM.

Embedded global navigation satellite system (GNSS) interference monitoring requires fast and memory-efficient inference to process large volumes of raw in-phase and quadrature (IQ) samples in real time. At the same time, increasingly expressive deep neural networks (DNNs) are needed for robust interference classification and characterization across diverse signal conditions. This creates a fundamental tension between predictive performance and deployability on resource-constrained hardware. In this paper, we investigate efficient network inference for GNSS interference characterization using iterative structured pruning, post-training static quantization, and hardware-aware zero-shot neural architecture search (NAS). Starting from MCUNet as a compact baseline, we analyze how model compression and automated architecture optimization affect model size, computational complexity, and memory usage while maintaining task performance. Experiments on a GNSS interference dataset, covering both classification and generalized characterization, show the benefits of combining compression and hardware-aware design for embedded deployment. Our results provide practical guidance for developing compact machine learning (ML) models for real-time GNSS interference monitoring on embedded platforms (iMXRT1062 MCU, Raspberry Pi Zero 2W, and Raspberry Pi 5).

Realtime omni-modal LMs support speech-centric conversations where users stream inputs, hear generated audio, and interrupt freely. Existing Omni-LM serving systems still rely on throughput-oriented LLM scheduling and LRU KV offloading. These policies ignore audio playback and multi-turn reuse: they may generate tokens far beyond what users hear, wasting work after barge-in, and evict KV state needed in the next turn. LiveServe is an interaction-aware serving system for realtime Omni-LM interaction. It exposes playback progress, speech activity, and barge-in events to the serving pipeline. The scheduler prioritizes first-audio and near-underrun sessions while limiting generation beyond the playback frontier. The KV manager uses next-use-aware eviction and preloads likely-needed KV during user speech to hide reload latency. On vLLM-Omni, LiveServe improves realtime serving across two Omni-LMs and mixed workloads. It lowers P90 audio TTFP by \(1.55\times\) on average and up to \(2.21\times\), while improving completed-request throughput by \(1.15\times\) on average and up to \(1.56\times\), and moves most KV reload work off the next-turn critical path.

While Diffusion Transformers (DiTs) have revolutionized high-fidelity video generation, their reliance on 3D full attention creates a quadratic computational bottleneck. Existing sparse methods face a dilemma: dynamic pruning suffers from prohibitive runtime overhead and memory fragmentation, while static heuristics fail to capture fine-grained dependencies. In this work, we propose ScalingAttention, a training-free framework grounded in a key inductive bias: while individual activations are input-dependent, the high-mass attention regions for each head rapidly converge to a stable, prompt-agnostic Intrinsic Sparse Topology. This topology is weight-encoded, scale-invariant, and efficient to extract. ScalingAttention decouples topology discovery from sparsity control via: (1) WEST (Weight-Encoded Sparse Topology), which extracts a robust block-sparse prior mask offline to eliminate runtime search; (2) FAST (Fidelity-Aware Sensitivity Tuning), which adaptively tunes head-wise sparsity based on diffusion fidelity requirements. To ensure practical acceleration, we co-design a hardware-aligned bit-wise block-sparse kernel. Experiments on Wan2.1 show up to 1.90X end-to-end speedup with superior fidelity, establishing a new Pareto frontier over state-of-the-art baselines.

Recent diffusion-based models have enabled realistic audio-driven avatar generation in real-time streaming. However, existing approaches struggle to maintain visual temporal consistency and fail to explicitly perceive user intent in complex interactive streaming scenarios. To address these challenges, we propose InteractiveAvatar, a real-time infinite-streaming video generation framework that supports visually consistent avatar video generation and intent-aware interactions. With autoregressive distillation, InteractiveAvatar achieves real-time str-eaming generation of human avatars over arbitrarily long durations. For visual consistency, we introduce a Long-Short Visual Memory (LSVM) mechanism that flexibly compresses historical visual information into compact tokens, preserving both short-range coherence and long-term consistency. To generate avatars with speeches and actions aligned with user intent, we propose a Reasoning-Reaction Module (RRM), which incorporates a State-Cycling strategy and a Cache-Switching mechanism. Extensive experimental results over diverse scenarios demonstrate that our method achieves state-of-the-art visual consistency in long-duration generation, while enabling complex user-avatar interaction in real time.

A-mode ultrasound (US) has emerged as a promising modality for hand and wrist motion tracking. Prior works have mainly addressed static gesture classification or regression of a few degrees of freedom (DoFs), typically relying on non-wearable systems and external computing devices, and highlight the need for strategies to ensure robustness to sensor repositioning. In this work, we propose a framework for robust whole-hand and wrist kinematic tracking via wearable A-mode US using the WULPUS platform, tackling the regression of 23 DoFs directly on the probe. First, we introduce a compact (11285 parameters) multi-output convolutional neural network combined with an incremental training strategy, which improves inter-session generalization and reduces mean absolute error by more than 17% compared to a non-incremental approach. Second, we demonstrate, for the first time, the feasibility of end-to-end hand and wrist kinematic tracking entirely on-device. We deploy the model on the WULPUS nRF52832 microcontroller, achieving 0.73 mJ per inference, 29.1 ms latency, and showing the feasibility of full operation (data acquisition, online inference, and BLE streaming of results) within 33 mW, enabling up to 36 hours of continuous use and an 88% reduction in wireless bandwidth compared to raw data transmission.

Localizing document tampering is extremely challenging, as manipulations are crafted to appear visually consistent and often leave only subtle traces that are nearly invisible to the human eye. In prior work, evaluation has been largely dominated by synthetic benchmarks that closely match the training distribution, and methods have shown steady progress under this setting. However, these gains often translate poorly to human-made forgeries and to cross-domain evaluation, where both the source documents and the tampering pipeline can change, leading to a distribution shift. In addition, since the introduction of the Frequency Perception Head for the discrete cosine transform (DCT) modality, it has become a standard choice, and subsequent work has largely focused on downstream modules and fusion strategies rather than revisiting the backbone itself. To help close this gap in cross-domain performance and improve the DCT backbone design, we propose \textbf{DiffNet}, a relatively simple yet effective RGB--DCT early-fusion architecture driven by two key design choices. First, to ensure that the decoder aggregates multi-scale inconsistency evidence rather than operating on raw, content-heavy activations, we apply a lightweight multi-level discrepancy transformation at the output of each backbone stage, replacing features with magnitude-only responses to learned zero-sum filters. Second, we design an efficient DCT-domain backbone that relies on a lightweight frequency-index-aware DCT--quantization joint embedding. Our approach achieves state-of-the-art performance on cross-domain and human-made document tampering localization, outperforming prior methods by around 30\%, with up to \(7\times\) higher throughput than the previous best model.

Modern video diffusion models achieve higher generation quality through scaling, but this also increases inference cost. Although many acceleration methods have been proposed, a central challenge is that the most effective acceleration strategy is highly instance-specific: a recipe that works well for one combination of model, hardware, and inference configuration often does not transfer to another. Different models vary in architecture, numerical sensitivity, and attention concentration patterns. Inference settings differ in spatial and temporal resolution and video duration, while hardware platforms differ in memory hierarchy, supported numerical formats, and kernel throughput. These factors create a large tuning space, making manual performance engineering costly. We present Sol Video Inference Engine, an agentic, native, training-free acceleration framework for video diffusion models. It organizes five broadly applicable techniques, cache, sparse attention, token pruning, quantization, and kernel fusion, into an agentic acceleration stack for instance-specific optimization. For a concrete deployment target defined by a model, hardware platform, and serving configuration, parallel skill agents optimize the implementation of each technique, an agent integrator composes them into a global acceleration stack, and a human validator provides feedback on generation quality. We instantiate this workflow on three video models with different sizes and architectures: 64B Cosmos3-Super, 22B LTX-2.3, and 2B SANA-Video. With little human effort, the full stack achieves more than 2x end-to-end acceleration while maintaining near-lossless VBench quality, demonstrating the effectiveness of the agent framework for video diffusion acceleration.

Mixture-of-Experts (MoE) models have become the de facto standard for scaling large language models. To maintain computational efficiency, modern MoE serving systems typically employ a hybrid parallelism strategy, combining Data Parallelism (DP) for attention stages with Expert Parallelism (EP) for MoE stages. However, this design necessitates frequent global synchronization barriers between attention DP groups and experts. In online serving, significant variance in request arrival rates and sequence lengths inherently leads to DP imbalance, causing severe synchronization stalls that degrade Time-to-First-Token (TTFT) and system throughput. We present ASAP, an asynchronous inference system specifically designed to accelerate the prefill phase of MoE models. ASAP disaggregates the attention and MoE stages and implements a fully asynchronous execution pipeline. This is achieved through a suite of specialized asynchronous communication primitives and four coordinated optimizations across request scheduling and model execution, which collectively dismantle global synchronization barriers. We implement and evaluate ASAP on CloudMatrix384 super-nodes, demonstrating that it improves SLO-compliant prefill throughput by 90% compared to state-of-the-art synchronous serving solutions.

Consistency distillation has significantly accelerated the inference of diffusion models. In this work, we reveal an intriguing asymmetry: while Logit-Normal sampling priors are highly efficacious for standard iterative generation, consistency distillation exhibits a distinctly different difficulty profile (e.g., U-shaped). We identify that the primary optimization bottlenecks reside at the boundary stages (initialization or final refinement) rather than the intermediate steps. To address the limitations of static sampling in accommodating evolving learning requirements, we propose Curvature-Adaptive Consistency Flow Matching (CACFM). By formulating distillation as a dynamic decision process, CACFM employs a lightweight Reinforcement Learning agent to actively probe Probability Flow ODE trajectories, automatically constructing an efficiency-oriented curriculum that prioritizes critical regions without manual scheduling. Integrated with a novel Flow Distribution Matching Distillation (DMD) objective, our approach achieves new state-of-the-art results on large-scale models such as FLUX and SDXL. It effectively mitigates structural deformities and preserves high-frequency details in extreme few-step regimes, achieving unprecedented visual fidelity.

The explosive demand for interactive Large Language Model serving has highlighted the management of the Key-Value cache's dynamic memory footprint as a critical area for performance optimization in inference engines. Modern inference systems overwhelmingly rely on time-centric scheduling heuristics, such as Shortest Job First. However, their theoretical optimality is rooted in traditional schedule modeling, failing to capture the highly dynamic, 2D spatio-temporal geometric growth specific to LLM inference mechanisms. To resolve this, we propose the geometry-aware online scheduling by introducing the Smallest Volume First (SVF) algorithm and its highly efficient variant, 1-bit SVF. Theoretically, we provide a rigorous mathematical foundation for our approach. Utilizing a novel proof methodology, we tighten the worst-case competitive ratio (\(\text{CR} \le 48 \rightarrow \text{CR} \le 5\)) for SVF with known output lengths. Building upon this core breakthrough, we complete a comprehensive theoretical taxonomy analyzing our algorithms across different traffic scenarios and information availability. Practically, we seamlessly integrate our approach as a plug-and-play layer in vLLM. Extensive evaluations on Llama-3.1 models demonstrate comprehensive performance gains: SVF delivers strong reductions in both average and tail latency, while 1-bit SVF, with merely a single bit information, achieves competitive throughput and latency. This work establishes a theoretically sound and empirically proven approach for resolving memory-constrained scheduling in modern LLM deployments. To facilitate future research, our code is available at https://github.com/Aurora-Kl/Geometry-Aware-Online-Scheduling.git.

In a tiered cache, eviction is a graph decision: removing one aligned storage block can disconnect downstream demand that never addressed that block directly, so request recency alone cannot price the action. This paper studies aligned eviction as a vertex-separation problem and gives a selection rule whose decisions carry independently checkable service-cut evidence. For every candidate block, it computes the exact weighted downstream demand cut, rejects actions that disconnect protected demand, and selects the minimum-impact admissible eviction. Reclamation is characterized as vertex separation: minimum-location reclamation reduces to node-capacitated flow, while minimum aligned block actions are NP-complete. In two-hop cache networks, one streaming pass evaluates every candidate impact; a matching adversarial construction proves that a history-only victim selector has unbounded one-step damage. The packet-scale implementation combines a seed-indexed exact-cardinality residency structure with collision-aware, 32-bank impact counters. Replay compression makes the result auditable: counter intervals reproduce the stream, exact monoid summaries retain every reported additive statistic, and a counting lower bound quantifies the state required by any exact all-candidate summary. A 144-scenario evaluation processes 582.90 trillion packets (404.86 PiB of simulated payload), validates the coordinate expectations, and exposes a zero-impact extreme-value transition near \(Nζ=\log m\). Complete impact vectors, decoded audit samples, telemetry, and logs remain within the ancillary-file budget. Finally, invalidation is monotone replicated state: fair asynchronous delivery converges without coordination, with a diameter bound under synchronous full-edge rounds. The architecture therefore binds capacity reclamation, path continuity, and distributed invalidation to one certifying interface.

Vision Transformers have achieved remarkable success in many fields, yet their deployment on edge devices remains challenging due to their substantial computational demands. Post-Training Quantization (PTQ) offers an attractive solution by compressing models using a small calibration set with minimal training overhead. However, most existing PTQ works adopt a static quantization paradigm that is uniformly applied to all instances. Given the substantial diversity of natural images, the activation distributions vary significantly across samples, making these methods inherently suboptimal. In this paper, we propose ScalePredictor, a dynamic quantization framework for accurate and efficient quantization scale learning of ViTs. We first reveal a hidden correlation between the distribution range of shallow-layer activations and the optimal scales of deeper layers. Based on this, we develop a scale learning mechanism that integrates an efficient range extraction approach to capture robust range statistics at the shallow stage, which are then fed into a Taylor-motivated polynomial scale projection module to generate all quantization scales simultaneously. With the efficiency of polynomial approximation, ScalePredictor introduces insignificant computational overhead while avoiding costly just-in-time calibration. Extensive experiments on ImageNet demonstrate that ScalePredictor consistently outperforms prior PTQ methods, achieving a more favorable accuracy-efficiency trade-off. Code and additional results are shown in the supplementary materials.

We propose Keyless Attention, an attention mechanism that eliminates the key projection entirely, operating over queries and values only. This yields a Value-Only Cache that reduces KV cache memory and access overhead by exactly 50% over standard attention, while matching or exceeding standard attention's decode throughput. Beyond efficiency, we introduce Depth-\(m\) Attention Factorization: standard attention computes a depth-2 factorization of the attention bilinear form, while Keyless Attention realizes a depth-\(m\) instance of this family. At m=3, Keyless Attention matches the projection matrix count of standard attention via a value-space routing matrix that replaces the key projection and introduces a coupling between routing and retrieval. Experiments across five models and four architectures (GPT-2 280M, GPT-2 557M, Pythia 410M, Qwen2 1.5B, and Llama 3.2 1B) show that Keyless Attention matches or outperforms standard QKV attention on perplexity in 4 out of 5 models. On downstream zero-shot evaluation (GPT-2 557M), Keyless Attention outperforms on 4 out of 5 commonsense reasoning benchmarks, while achieving 50% KV cache reduction throughout.

Modern Mixture-of-Experts (MoE) models place most of their parameters in expert layers, yet only a small fraction of those experts are used for any token. The unused weights must still be stored where the GPU can reach them. On commodity GPUs the common fix is layer-level CPU offloading, which keeps memory low but streams all of a layer's experts across PCIe on every forward pass, losing much of MoE's sparsity benefit. We cast low-resource MoE serving as a working-set management problem on the GPU: routed expert weights and the key-value (KV) cache are two streams of memory demand competing for limited VRAM. We realize this in WiSP (Working-Set Paging), a routing-aware expert pager that plugs into an unmodified serving engine with byte-identical outputs. Keeping resident only the experts a workload reuses, WiSP reaches up to 1.95x the decode throughput of static offload at the same memory budget when the model does not fit. We also find that prefetching experts from predicted routing helps little in single-stream decode: the bottleneck is PCIe bandwidth, not prediction accuracy. This shifts the question from prefetching to allocation: how should VRAM be split between experts and the KV cache? We answer with MV-WSA (Marginal-Value Working-Set Allocation), which equalizes marginal latency benefit per byte subject to a KV admission floor. MV-WSA runs either as an offline configurator or as an online controller that resizes both pools while serving. In real serving the offline configurator is the only policy we test that does well on both prefill and decode; in trace-driven simulation it stays within a few percent of a per-workflow oracle while fixed splits are about 20% worse. The online controller adds a further 1.20x without changing model outputs.