Add to Chrome
Add to Firefox
Add to EdgeRelay Buffer Independent Communication over Pooled HBM for Efficient MoE Inference on Ascend
May 07, 2026
Mixture-of-Experts (MoE) inference requires large-scale token exchange across devices, making dispatch and combine major bottlenecks in both prefill and decode. Beyond network transfer, routing-driven layout transformation, temporary relay, and output restoration can add substantial overhead. Existing MoE communication paths are often buffer-centric, using explicit inter-process relay and reordering buffers around collective transfer. This report presents a relay-buffer-free communication design for MoE inference acceleration on Ascend systems. The design reorganizes dispatch and combine around direct placement into destination expert windows and direct reading from remote expert windows. Built on globally pooled high-bandwidth memory and symmetric-memory allocation, it removes most intermediate relay and reordering buffers while retaining only lightweight control state, including counts, offsets, and synchronization metadata. We instantiate the design as two schedules for the main phases of MoE inference: a prefill schedule with richer planning state for throughput-oriented execution, and a compact decode schedule for latency-sensitive execution. Experiments on Ascend-based MoE workloads show reduced dispatch and combine latency in both settings. At the serving level, the implementation improves time to first token (TTFT), preserves competitive time per output token (TPOT), and enlarges the feasible scheduling space under practical latency constraints. These results indicate that, on platforms with globally addressable device memory, reducing intermediate buffering and output restoration around expert execution is an effective direction for accelerating MoE inference.
FitText: Evolving Agent Tool Ecologies via Memetic Retrieval
May 04, 2026A semantic gap separates how users describe tasks from how tools are documented. As API ecosystems scale to tens of thousands of endpoints, static retrieval from the initial query alone cannot bridge this gap: the agent's understanding of what it needs evolves during execution, but its tool set does not. We introduce FitText, a training-free framework that makes retrieval dynamic by embedding it directly in the agent's reasoning loop. FitText generates natural-language pseudo-tool descriptions as retrieval probes, refines them iteratively using retrieval feedback, and explores diverse alternatives through stochastic generation. Memetic Retrieval adds evolutionary selection pressure over candidate descriptions, guided by a tool memory that avoids redundant search. On ToolRet (43k tools, 4 domains), FitText improves average retrieval rank from 8.81 to 2.78; on StableToolBench (16,464 APIs), it achieves a 0.73 average pass rate--a 24-point absolute gain over static query retrieval. The gains transfer across base models capable of acting as competent semantic operators; under weaker base models, Memetic's evolutionary search inverts--amplifying noise rather than refining signal--surfacing model capacity as a prerequisite for evolutionary tool exploration.
HeavySkill: Heavy Thinking as the Inner Skill in Agentic Harness
May 04, 2026Recent advances in agentic harness with orchestration frameworks that coordinate multiple agents with memory, skills, and tool use have achieved remarkable success in complex reasoning tasks. However, the underlying mechanism that truly drives performance remains obscured behind intricate system designs. In this paper, we propose HeavySkill, a perspective that views heavy thinking not only as a minimal execution unit in orchestration harness but also as an inner skill internalized within the model's parameters that drives the orchestrator to solve complex tasks. We identify this skill as a two-stage pipeline, i.e., parallel reasoning then summarization, which can operate beneath any agentic harness. We present a systematic empirical study of HeavySkill across diverse domains. Our results show that this inner skill consistently outperforms traditional Best-of-N (BoN) strategies; notably, stronger LLMs can even approach Pass@N performance. Crucially, we demonstrate that the depth and width of heavy thinking, as a learnable skill, can be further scaled via reinforcement learning, offering a promising path toward self-evolving LLMs that internalize complex reasoning without relying on brittle orchestration layers.
Geometric and Spectral Alignment for Deep Neural Network II
May 04, 2026This paper develops the angular and static-channel component of Geometric and Spectral Alignment for residual Jacobian chains. Starting from Cartan-coordinate rigidity and fitted effective-rank windows, we study how dominant singular subspaces are transported across adjacent layers and how the resulting finite matrices can be displayed in physical channel coordinates. The main results are deterministic, margin-verified results. We bound the error between full interface transport and its dominant-window truncation, add fitted-tail errors so that empirical spectra can be certified against the Gibbs--Cartan tail model, and distinguish source-mode incidence from fully physical input-output channel incidence. Given row groups and active supports, the Physical Alignment Matrix decomposes orthogonally as core plus overlap plus noise. Active-column gaps, pairwise overlap margins, and noise bounds combine into a static certificate radius under which the full transport and the truncated transport induce the same active supports, pairwise incidence graph, SRS sets, hub columns, and core/overlap/noise masks. The finer SC/SA/ST labels of the Invariant Channel Mapping require additional row-energy and profile-correlation margins, stated as explicit perturbation tests. The empirical section reports the matrices and block-energy heatmaps that measure these certificate quantities across CNNs, language models, and vision/diffusion backbones. The figures are interpreted as finite-dimensional measurements; complete membership in the Physical GSA certificate domain requires checking the numerical margin protocol stated in Section 10.
Progressive Approximation in Deep Residual Networks: Theory and Validation
Apr 27, 2026The Universal Approximation Theorem (UAT) guarantees universal function approximation but does not explain how residual models distribute approximation across layers. We reframe residual networks as a layer-wise approximation process that builds an approximation trajectory from input to target, and prove the existence of progressive trajectories where error decreases monotonically with depth. It reveals that residual networks can implement structured, step-by-step refinement rather than end-to-end (E2E) black-box mapping. Building on this, we propose Layer-wise Progressive Approximation (LPA), a theoretically grounded training principle that explicitly aligns each layer with its residual target to realize such trajectories. LPA is architecture-agnostic: we observe progressive behavior in residual FNNs, ResNets, and Transformers across tasks including complex surface fitting, image classification, and NLP with LLMs for generation and classification. Crucially, this enables ``train once, use $N$ models": a single network yields useful predictions at every depth, supporting efficient shallow inference without retraining. Our work unifies approximation theory with practical deep learning, providing a new lens on representation learning and a flexible framework for multi-depth deployment. The source code will be released unpon acceptance at https://(open\_upon\_acceptance).
Sub-Token Routing in LoRA for Adaptation and Query-Aware KV Compression
Apr 23, 2026Sub-token routing offers a finer control axis for transformer efficiency than the coarse units used in most prior work, such as tokens, pages, heads, or layers. In this paper, we study routing within a token representation itself in LoRA-adapted transformers. The motivation is that a relevant token need not be internally uniform: under a retention budget, preserved value groups are distributed unevenly both across tokens and within tokens, which suggests that KV compression need not be an all-or-nothing decision at token level. We study this fine-grained routing mechanism in two settings. For compression-aware language modeling, we introduce a query-independent design that combines routed subspace LoRA with value-group routing on the KV path. For downstream-task-preserving KV compression, we introduce a query-aware design in which a predictor-based selector allocates a global retention budget over context-token/value-group pairs using query-conditioned relevance. Experiments show that the query-independent design improves the quality-compression tradeoff for language modeling, while the query-aware design preserves downstream behavior under reduced KV budgets. We further examine the relation between token-level and sub-token-level query-aware routing, and show that they form complementary compression axes: token-level methods determine which tokens survive globally, while sub-token routing determines how the surviving tokens are compressed internally.
Open-H-Embodiment: A Large-Scale Dataset for Enabling Foundation Models in Medical Robotics
Apr 22, 2026Autonomous medical robots hold promise to improve patient outcomes, reduce provider workload, democratize access to care, and enable superhuman precision. However, autonomous medical robotics has been limited by a fundamental data problem: existing medical robotic datasets are small, single-embodiment, and rarely shared openly, restricting the development of foundation models that the field needs to advance. We introduce Open-H-Embodiment, the largest open dataset of medical robotic video with synchronized kinematics to date, spanning more than 49 institutions and multiple robotic platforms including the CMR Versius, Intuitive Surgical's da Vinci, da Vinci Research Kit (dVRK), Rob Surgical BiTrack, Virtual Incision's MIRA, Moon Surgical Maestro, and a variety of custom systems, spanning surgical manipulation, robotic ultrasound, and endoscopy procedures. We demonstrate the research enabled by this dataset through two foundation models. GR00T-H is the first open foundation vision-language-action model for medical robotics, which is the only evaluated model to achieve full end-to-end task completion on a structured suturing benchmark (25% of trials vs. 0% for all others) and achieves 64% average success across a 29-step ex vivo suturing sequence. We also train Cosmos-H-Surgical-Simulator, the first action-conditioned world model to enable multi-embodiment surgical simulation from a single checkpoint, spanning nine robotic platforms and supporting in silico policy evaluation and synthetic data generation for the medical domain. These results suggest that open, large-scale medical robot data collection can serve as critical infrastructure for the research community, enabling advances in robot learning, world modeling, and beyond.
Fast-then-Fine: A Two-Stage Framework with Multi-Granular Representation for Cross-Modal Retrieval in Remote Sensing
Apr 22, 2026Remote sensing (RS) image-text retrieval plays a critical role in understanding massive RS imagery. However, the dense multi-object distribution and complex backgrounds in RS imagery make it difficult to simultaneously achieve fine-grained cross-modal alignment and efficient retrieval. Existing methods either rely on complex cross-modal interactions that lead to low retrieval efficiency, or depend on large-scale vision-language model pre-training, which requires massive data and computational resources. To address these issues, we propose a fast-then-fine (FTF) two-stage retrieval framework that decomposes retrieval into a text-agnostic recall stage for efficient candidate selection and a text-guided rerank stage for fine-grained alignment. Specifically, in the recall stage, text-agnostic coarse-grained representations are employed for efficient candidate selection; in the rerank stage, a parameter-free balanced text-guided interaction block enhances fine-grained alignment without introducing additional learnable parameters. Furthermore, an inter- and intra-modal loss is designed to jointly optimize cross-modal alignment across multi-granular representations. Extensive experiments on public benchmarks demonstrate that the FTF achieves competitive retrieval accuracy while significantly improving retrieval efficiency compared with existing methods.
Well Begun is Half Done: Training-Free and Model-Agnostic Semantically Guaranteed User Representation Initialization for Multimodal Recommendation
Apr 16, 2026Recent advancements in multimodal recommendations, which leverage diverse modality information to mitigate data sparsity and improve recommendation accuracy, have gained significant attention. However, existing multimodal recommendations overlook the critical role of user representation initialization. Unlike items, which are naturally associated with rich modality information, users lack such inherent information. Consequently, item representations initialized based on meaningful modality information and user representations initialized randomly exhibit a significant semantic gap. To this end, we propose a Semantically Guaranteed User Representation Initialization (SG-URInit). SG-URInit constructs the initial representation for each user by integrating both the modality features of the items they have interacted with and the global features of their corresponding clusters. SG-URInit enables the initialization of semantically enriched user representations that effectively capture both local (item-level) and global (cluster-level) semantics. Our SG-URInit is training-free and model-agnostic, meaning it can be seamlessly integrated into existing multimodal recommendation models without incurring any additional computational overhead during training. Extensive experiments on multiple real-world datasets demonstrate that incorporating SG-URInit into advanced multimodal recommendation models significantly enhances recommendation performance. Furthermore, the results show that SG-URInit can further alleviate the item cold-start problem and also accelerate model convergence, making it an efficient and practical solution for multimodal recommendations.
RoSLAC: Robust Simultaneous Localization and Calibration of Multiple Magnetometers
Apr 15, 2026Localization of autonomous mobile robots (AMRs) in enclosed or semi-enclosed environments such as offices, hotels, hospitals, indoor parking facilities, and underground spaces where GPS signals are weak or unavailable remains a major obstacle to the deployment of fully autonomous systems. Infrastructure-based localization approaches, such as QR codes and RFID, are constrained by high installation and maintenance costs as well as limited flexibility, while onboard sensor-based methods, including LiDAR- and vision-based solutions, are affected by ambiguous geometric features and frequent occlusions caused by dynamic obstacles such as pedestrians. Ambient magnetic field (AMF)-based localization has therefore attracted growing interest in recent years because it does not rely on external infrastructure or geometric features, making it well-suited for AMR applications such as service robots and security robots. However, magnetometer measurements are often corrupted by distortions caused by ferromagnetic materials present on the sensor platform, which bias the AMF and degrade localization reliability. As a result, accurate magnetometer calibration to estimate distortion parameters becomes essential. Conventional calibration methods that rely on rotating the magnetometer are impractical for large and heavy platforms. To address this limitation, this paper proposes a robust simultaneous localization and calibration (RoSLAC) approach based on alternating optimization, which iteratively and efficiently estimates both the platform pose and magnetometer calibration parameters. Extensive evaluations conducted in high-fidelity simulation and real-world environments demonstrate that the proposed RoSLAC method achieves high localization accuracy while maintaining low computational cost compared with state-of-the-art magnetometer calibration techniques.