arXiv subjects

Fan Yang

Publications and source records attributed to Fan Yang.

At least 55 records · Page 3Linked to original sources

arXiv2026

Joint Branch-Space Transform Coding for Diffusion Activation Quantization with Classifier-Free Guidance

Post-training quantization for diffusion models increasingly exploits timestep, feature, and layer structure. While recent work has begun incorporating CFG structure into diffusion quantization, activation quantization still operates independently across conditional and unconditional coordinates, leaving cross-activation structure unexploited. We show that matched CFG activations form a strongly correlated two-dimensional source and that, under a fixed bit budget, the choice of branch coding basis materially affects quantization fidelity. Motivated by this observation, we introduce branch-space transform coding, which rotates matched CFG branches via an offline derived 2x2 orthogonal matrix, requiring minimal modifications to model parameters or the quantization pipeline. We further derive the Guidance-Correlation Branch Transform (GCBT), which jointly incorporates the CFG guidance direction and cross-branch second moments. Under an equal-rate quantization-noise surrogate, GCBT admits a closed-form per-layer solution without gradient optimization or angle search. Applied on top of existing diffusion PTQ methods, GCBT yields statistically significant fidelity gains in most evaluated comparisons with no statistically significant degradation, while leaving the underlying host quantization pipeline unchanged.

cs.CV↗

arXiv2026

Breakdown of material-time scaling in the microscopic dynamics of metallic glasses

Physical aging in glasses is often described using the concept of material time, which assumes the underlying relaxation mechanism to remain the same as in equilibrium, but with a rate that changes during aging. Whether material time also captures the microscopic dynamics of metallic glasses, which differ qualitatively between the glass and the equilibrium supercooled liquid, is unclear. Using simultaneous time-resolved X-ray diffraction and X-ray photon-correlation spectroscopy, we follow the full equilibration of a metallic glass after small temperature jumps starting from equilibrium. The time evolution of the first sharp diffraction peak position exhibits the asymmetry between temperature up- and down-jumps characteristic of material-time aging and is quantitatively described by the Tool-Narayanaswamy-Moynihan model, which is based on this concept. By contrast, the microscopic density-density autocorrelation functions strongly violate material-time scaling. Compressed-exponential relaxation emerges even after temperature changes as small as 4K, irrespective of the direction of the temperature jump, and continuously evolves into a stretched-exponential relaxation shape following the equilibration of the average structure. This behavior is consistent with a transient stress-mediated contribution associated with volumetric frustration that disappears upon equilibration. Our results show that a material time can describe the evolution of average structural properties even when the corresponding microscopic dynamics do not obey material-time scaling.

cond-mat.dis-nn↗

arXiv2026

MoSE: Mode-Switching Expander for Mixed LLM Training and Inference

AI clusters increasingly run large language model (LLM) inference and training on the same fabric. Prefill-decode (P-D) disaggregation creates key-value (KV) cache transfers between prefill and decode groups, whereas training collectives and all-to-all traffic benefit from near-uniform global connectivity. A static sparse topology can therefore be poorly matched to one of the two traffic patterns. We present Mode-Switching Expander (MoSE), a reconfigurable expander that treats topology design as a fixed-degree edge-allocation problem. MoSE reallocates the same sparse edge budget toward direct P-D connectivity in inference-heavy modes and restores a uniform random regular expander in training-heavy modes. We evaluate MoSE using a 1024-group flow-level topology model, shortest-path routing, and two mixed workloads. Across 20 seeds, MoSE reduces average and 95th-percentile (P95) load-aware KV communication cost by 90.8\% and 91.9\% relative to Static-Training in the inference-heavy mode. In the training-heavy mode, it reduces average and P95 training communication cost by 22.7\% and 27.6\% relative to stale Static-Inference. These results show that coarse-grained topology switching can support both workload modes without additional ports or routing changes.

cs.NI↗

arXiv2026

Spatio-Temporal Wireless-Optical Planning for Multi-UAV Networks

Multi-unmanned aerial vehicle (UAV) networks in urban low-altitude environments couple UAV mobility, wireless access, and optical backhaul resources. Existing path-planning methods optimize flight distance or wireless signal quality, but can still concentrate traffic on shared optical backhaul links. We present Spatio-Temporal Wireless-Optical (STWO) planner, a backhaul-aware path-planning algorithm that jointly considers flight distance, wireless link quality, and time-varying optical-link offered-load ratio. STWO updates backhaul occupancy during sequential multi-UAV planning, allowing later UAVs to avoid congested optical paths while maintaining wireless connectivity. Experiments show that STWO reduces peak optical-link offered-load ratio by up to 56.4\% and congestion ratio by up to 72.6\% under dense UAV deployment, demonstrating the importance of wireless-optical awareness for reliable multi-UAV transmission.

cs.NI↗

arXiv2026

KwaiMind Technical Report

Commercial image editing requires product identity preservation, accurate text rendering, and user appeal alongside general editing quality. We present KwaiMind, an image editing system combining general capabilities with e-commerce specialization. An agent-based data engine maintains approximately 1.8 million high-quality editing pairs. Built on a multimodal diffusion transformer, KwaiMind undergoes continued pre-training and supervised fine-tuning, followed by preference optimization and online reinforcement learning. A general-purpose vision-language judge and specialized rewards for click-through rate (CTR), text rendering, and product consistency guide specialized policies, which are consolidated through on-policy distillation. We introduce Ecom-Bench, covering 11 commercial editing tasks with task-specific visual evaluation and CTR-based ranking. KwaiMind achieves the strongest overall scores among evaluated open-source editors on ImgEdit, GEdit, both language splits of REDEdit, and Ecom-Bench visual quality, and the highest aggregate CTR ranking score among compared systems. Offline, CTR-guided optimization increases the proportion of generated images whose predicted CTR exceeds that of the original product image from 12.16% to 37.41%. In an online A/B experiment, CTR-based selection of product main images yields an approximately 2.44% relative increase in actual CTR. These results demonstrate the value of domain-specific data and reward-driven alignment for commercial image editing.

cs.CV↗

arXiv2026

RingStitch: Demand-Aware Optical Stitching for Fragmented TPU Clusters

Large-scale AI training clusters increasingly use optical circuit switching (OCS) to reconfigure rack-level interconnects and create elastic accelerator slices. In multi-tenant TPU-style clusters, however, small and medium jobs often leave partial free capacity stranded inside racks. Although the aggregate free capacity may be sufficient for a new job, it cannot be used by local placement or coarse full-rack stitching. This paper presents RingStitch, an OCS-based defragmentation scheduler that turns fragmented rack capacity into schedulable resources. RingStitch follows a local-first policy, stitches compact cross-rack fragments only when needed, and orders the selected racks into a low-cost logical ring. Simulations on a TPU 8t-like SuperPod model show that RingStitch improves schedulability over local and full-rack baselines.

cs.NI↗

arXiv2026

Bit-Accurate Modeling of GPU Matrix Multiply-Accumulate Units: Demystifying Numerical Discrepancy and Accuracy

Modern AI accelerators rely on matrix multiply-accumulate units (MMAUs), such as NVIDIA Tensor Cores and AMD Matrix Cores, to accelerate deep neural network workloads. MMAUs expose only instruction-level or API-level interfaces of matrix multiply-accumulate (MMA) operations, while leaving internal floating-point arithmetic behavior undocumented. Consequently, MMAUs across vendors and architectural generations often produce numerical discrepancies for identical inputs, and sometimes exhibit reduced numerical accuracy that can cause training instability. Diagnosing and understanding the root causes of these effects is challenging without white-box models of their arithmetic behavior. This paper proposes closed-loop feature probing (CLFP), a generic and systematic framework for constructing bit-accurate arithmetic behavior models of MMA operations. Based on this framework, we analyze all MMA instructions on ten GPU architectures spanning NVIDIA Volta through RTX Blackwell and AMD CDNA1 through CDNA3, and derive the first bit-accurate arithmetic models for these MMAUs. Our models explain previously observed cross-platform numerical discrepancies and accuracy issues, enable white-box numerical error analysis, reveal four types of precision bottlenecks and one type of numerical asymmetry, and inform software workarounds as well as design suggestions for future MMAUs. This work is open-source at https://github.com/microsoft/MMA-Sim

cs.AR↗

arXiv2026

LongCat-DeepResearch Technical Report

We present LongCat-DeepResearch, a deep research system that combines an enhanced LongCat model with a multi-agent workflow for producing comprehensive, evidence-grounded reports. The workflow separates global planning from detailed investigation and coordinates revision at the section level. Multiple planning agents first explore external sources and refine an actionable research plan, termed ResearchSpec. Research agents then investigate and draft their assigned sections in parallel, gathering additional evidence in separate contexts as their analyses develop. Once the sections are assembled, global review guides targeted local revisions, reducing reliance on repeated full-report rewriting. This workflow also supports the construction of research tasks and trajectories for the mid-training and post-training of LongCat's general-purpose models. LongCat-DeepResearch achieves 55.25 on DeepResearchBench, 51.35 on DeepResearchBench II, and 79.83 on ResearchRubrics. On an in-house benchmark, it scores 76.04, ranking second among four compared systems. Development-set analyses show benefits from combining planning perspectives, while further planning refinement has mixed effects. Additional editing improves average automatic readability preference across two benchmarks, with different trends on each.

cs.AI↗

arXiv2026

Reason What Matters: Retrieval-Grounded Reasoning for Universal Multimodal Embeddings

Universal multimodal embedding (UME) maps multimodal inputs into a shared embedding space for diverse retrieval tasks. Recent methods improve embeddings through Chain-of-Thought (CoT) reasoning optimized with GRPO using retrieval rewards. However, existing methods overlook the mismatch bettween candidate-aware retrieval supervision and input-only CoT generation: (1)trajectory-level rewards convey retrieval outcomes without explicitly identifying the input-supported evidence that distinguishes the positive from hard negatives; (2) input-only generation cannot directly assess whether further reasoning improves retrieval, potentially producing redundant CoTs with substantial latency. To bridge this gap, we propose Reason What Matters (ReWAM), a retrieval-grounded framework that aligns candidate-aware supervision with input-only generation. Specifically, we introduce Retrieval-Aware Self-Distillation (RASD), which extracts privileged guidance from input-supported facts and evidence distinguishing the positive from hard negatives. Conditioned on this guidance, an on-policy self-teacher provides token-level feedback to refine credit assignment, directing policy updates toward retrieval-relevant reasoning grounded in the input. We further propose Retrieval-Adaptive Inference (RAI), which learns a retrieval-aware stopping criterion from prefix-level retrieval feedback. It stops redundant reasoning without candidate access and uses speculative decoding to further reduce CoT latency. Extensive experiments on MMEB-V2 and MRMR demonstrate that ReWAM achieves state-of-the-art retrieval performance while delivering up to 5x the inference throughput of competitive explicit-CoT UME methods. ReWAM thus enables high-quality retrieval through efficient input-only reasoning, making explicit CoT practical for corpus-scale multimodal retrieval. The code will be publicly available.

cs.AI↗

arXiv2026

PRQuant: Permutation Residual Quantization for Low-Overhead Inference

Low-bit quantization of linear layers is often dominated by a small number of outlier channels. Existing smoothing, rotation, and residual-based methods can mitigate this issue, but may shift the quantization bottleneck to weights or introduce costly online operations. To address these limitations, we propose PRQuant (Permutation Residual Quantization), a training-free framework that combines channel permutation with offline weight residual compensation. PRQuant identifies the scaled-weight columns with the largest quantization errors and permutes them into contiguous tail blocks. This structure allows the corresponding weight residuals to be precomputed entirely offline, while replacing scattered activation gathering with simple contiguous access during inference, yielding a single regular MXFP4 GEMM for compensated computation. Experiments show that PRQuant substantially reduces down-projection reconstruction error, with scaling and residual compensation providing the main numerical gains while permutation enables a hardware-friendly contiguous layout. Comprehensive experiment results on Qwen3-4B-Instruct-2507 and Qwen3-30B-A3B-Instruct-2507 illustrate that PRQuant achieves up to averagelly 2.6x and 1.8x operator speedup over BF16 respectively, while preserving near plain MXFP4 end-to-end decoding efficiency. Across five downstream benchmarks, PRQuant achieves the best average accuracy among the quantized methods, improving accuracy over MXFP4 by 1.24 and 0.55, respectively.

cs.LG↗

arXiv2026

CaC: Advancing Video Reward Models via Hierarchical Spatiotemporal Concentrating

In this paper, we propose Concentrate and Concentrate (CaC), a coarse-to-fine anomaly reward model based on Vision-Language Models. During inference, it first conducts a global temporal scan to anchor anomalous time windows, then performs fine-grained spatial grounding within the localized interval, and finally derives robust judgments via structured spatiotemporal Chain-of-Thought reasoning. To equip the model with these capabilities, we construct the first large-scale generated video anomaly dataset with per-frame bounding-box annotations, temporal anomaly windows, and fine-grained attribution labels. Building on this dataset, we design a three-stage progressive training paradigm. The model initially learns spatial and temporal anchoring through single- and multi-frame supervised fine-tuning, and then is optimized by a reinforcement learning strategy based on two-turn Group Relative Policy Optimization (GRPO). Beyond conventional accuracy rewards, we introduce Temporal and Spatial IoU rewards to supervise the intermediate localization process, effectively guiding the model toward more grounded and interpretable spatiotemporal reasoning. Extensive experiments demonstrate that CaC can stably concentrate on subtle anomalies, achieving a 25.7% accuracy improvement on fine-grained anomaly benchmarks and, when used as a reward signal, CaC reduces generated-video anomalies by 11.7% while improving overall video quality.

cs.CV↗

arXiv2026

Tool-Augmented On-Policy Distillation for LLM Domain Adaptation in Sequence-Based Omics Tasks

Multi-omics sequences contain complex biological patterns, yet deciphering their mechanisms for automated scientific discovery remains challenging. As large language models (LLMs) interpret these sequences, evaluating both predictions and scientific reasoning is critical. However, existing benchmarks for multi-omics sequence tasks rely on classification and regression metrics, neglecting whether models grasp the underlying biological evidence. We introduce OmicsBench, the first reasoning benchmark for multi-omics sequences, comprising 1,160 expert-validated questions across six tasks spanning DNA regulation, RNA processing, and protein function. OmicsBench requires traceable evidence chains, evaluated using instance-specific rubrics developed with domain experts. Evaluating 17 LLMs reveals an inverse relationship: while scientific LLMs outperform general-purpose LLMs in sequence classification accuracy, they fail to provide valid evidence to support their predictions. One plausible interpretation is shortcut learning: specialized models may rely on statistical patterns rather than the biological mechanisms needed for scientific discovery. Motivated by this finding, we introduce tool-augmented on-policy distillation (TA-OPD), a post-training method to align sequence prediction with evidence-grounded biological reasoning. Across five Qwen3.5 models spanning 0.8B to 27B parameters, TA-OPD consistently strengthens biological evidence grounding while improving predictive performance on most tasks. These gains persist across model scales, indicating that stronger sequence reasoning does not arise solely from increased model capacity, but can be improved through evidence-aware training. Together, OmicsBench and TA-OPD provide a framework for diagnosing reasoning failures in multi-omics LLMs and a path toward models whose predictions are better grounded in biologically meaningful evidence.

cs.LG↗

arXiv2026

Small circumference in regular sublinear expanders

Sublinear expansion is weak enough to be extracted from arbitrary graphs while retaining nearly all of their average degree, yet it has proved strong enough to force global structures in many sparse extremal problems. Letzter, Methuku and Sudakov [JLMS 2026] proved the existence of nearly Hamilton cycles in sufficiently dense regular sublinear expanders. Montgomery [ICM 2026] subsequently conjectured that, every $d$-regular sublinear expander with $d$ sufficiently large (but constant) is Hamiltonian. We disprove this conjecture in a strong form by constructing $n$-vertex $d$-regular sublinear expanders with $d=\left(\frac12+o(1)\right)\log^2 n$, which can forbid any cycle covering an arbitrarily small given positive constant portion of its vertices. The construction blows up one side of a biregular Ramanujan graph into almost-complete blocks and keeps the other side as a sparse vertex separator. We also prove a similar statement for a closely related notion of edge expanders. For every sufficiently small $γ>0$, there is an infinite family of $n$-vertex $d$-regular $γ$-edge-expanders with $d=Θ(γ^{-1})$ and circumference $O(γn)$, matching the standard lower bound $Ω(γn)$. The construction also comes from an expanding regular core such that each vertex has an almost-complete graph attached to it. The core guarantees edge expansion, while the single-vertex attachments confine every cycle.

math.CO↗

arXiv2026

Efficient Simulation of Nonreciprocal Many-body Physics via Quantum Feedback

We present a scheme to efficiently simulate nonreciprocal many-body spin models using a non-Markovian open system. Our scheme utilizes a single quantum emitter coupled to a waveguide mode that is fed back to the same emitter after a time delay. With this coherent delayed feedback, an effective nonreciprocal interaction can be engineered, wherein the emitter at earlier times affects itself at later times, creating a scalable spin chain. We show that all the steady-state quantities of such spin models can be accessed through measurement of the emitter and the output field. Furthermore, using classical feedback that resets the emitter, quench dynamics from arbitrary product states can be simulated. We demonstrate striking features of nonreciprocal models, such as the Liouvillian skin effect, anomalous relaxation dynamics, and quasi-long-range order of output photons. Finally, we show that the proposal is amenable to experimental realization and robust to imperfections. Our work establishes a feasible way to scale up a nonreciprocal many-body system, and provides a new route towards generation of exotic states of light.

quant-ph↗

arXiv2026

Value-Sensitive Delegation in Everyday AI Agent Use: Evidence from OpenClaw

Users increasingly delegate work to autonomous AI agents, yet evaluations typically measure task completion rather than the values users prioritize. Using Value Sensitive Design, we analyzed, with LLM assistance, 73,093 first-person Reddit posts about using OpenClaw, each for its human value, agent aspect, value fulfillment, and user outcome. The 21 values form six value groups, including Autonomous, Dependable, and Affordable Operation, Bounded Reach, Reviewability, and Equitable Access. Relative to each aspect's corpus share, values clustered not at the agent's outputs but at the operating conditions users set around a run. Values were usually met where users described what the agent delivered, in five of six groups, and mostly unmet where users described supervising it, in all six groups. We conceptualize this pattern as value-sensitive delegation. Supporting human values requires attention not only to what an agent accomplishes, but to the conditions users set around delegation, including cost, access, and oversight.

cs.HC↗

arXiv2026

On 2-distance-transitive circulant digraphs

Circulant digraphs form a prominent class of Cayley digraphs defined on finite cyclic groups. Building on the existing classification of $2$-arc-transitive circulant graphs, this paper presents a complete classification of $2$-distance-transitive circulant digraphs. Our main theorem establishes that every connected $2$-distance-transitive circulant digraph is isomorphic to one of the following: the undirected cycle $C_n$, the complete bipartite graph $\K_{\frac{n}{2},\frac{n}{2}}$, the complete multipartite graph $\K_{m[b]}$ with $m\geq 3,b\geq 2$, the graph $\K_{\frac{n}{2},\frac{n}{2}}-\frac{n}{2}\K_2$ for odd $\frac{n}{2}$, prime-order Paley graphs, the directed cycle $\overrightarrow{C}_n$, the oriented graph $G(p^m,r)$ satisfying Condition~\ref{p-power-normal-2dt-cond}, the oriented graph $ C_r(b,1)$ with $r\geq 3,b\geq 2$ and $rb=n$, the lexicographic product oriented graph $ G(p^m,r)[\overline{\K}_d]$ where $G(p^m,r)$ obeys Condition~\ref{p-power-normal-2dt-cond}.

math.CO↗

arXiv2026

Online Multimodal Workload Assessment in Contact-Rich Physical Human-Robot Interaction

Contact-rich physical human--robot interaction (pHRI) imposes time-varying demands associated with physical interaction, motor regulation, and physiological response, motivating continuous assessment of interaction workload. This paper presents an online multimodal assessment framework that integrates interaction wrench, planar tool-center-point (TCP) kinematics, and skin conductance level (SCL) into four interpretable workload-related factors. Their relative contributions are adjusted using path curvature to reflect changes in motion demand and task progression to account for gradual physiological variation over time. The framework was evaluated with 24 participants across 18 controlled combinations of temperature, acoustic noise, and illuminance under two admittance-control modes. Strict leave-one-subject-out (LOSO) evaluation used standardized pupil diameter ($\mathrm{PD}_z$) as an independent physiological reference and included comparisons with static variants and representative state-of-the-art learning-based baselines. The proposed framework achieves a cohort-mean $30\,\mathrm{s}$ block-wise Spearman correlation of $ρ_{30}=0.308$ with the physiological reference, with positive subject-level correspondence in 23 of 24 participants. Its overall performance is comparable to the state-of-the-art learning-based baseline. At the same time, our framework keeps the assessment process transparent through explicit workload-related factors and defined weighting rules, while outperforming the corresponding fixed-weight formulation. The framework also maintains consistent performance across the two tested admittance-control modes. These results support a transparent and interpretable approach to continuous interaction workload assessment in contact-rich pHRI.

cs.RO↗

arXiv2026

CARES: A Conversational AI System for Regulation-Grounded Safety Reporting in Construction Education

Construction safety reporting often relies on manual logs and static templates that provide limited feedback and leave daily activities disconnected from relevant regulations. This paper introduces CARES (Conversational AI Reporting for Enhanced Safety), a conversational AI system that integrates regulatory guidance into daily reporting to support construction safety education. CARES combines proactive multi-agent dialogue, retrieval-augmented generation (RAG), and automated report generation. The system guides users through reporting tasks, retrieves relevant regulatory passages using hybrid retrieval, and converts conversations into structured daily reports. Regulatory sources and the evolving report are displayed alongside the dialogue to support user review and correction. A preliminary evaluation involving 15 construction management students assessed retrieval quality, response faithfulness, and conversational relevance. CARES achieved an overall faithfulness score of 0.74 and an answer relevance score of 1.00, while initial retrieval ranking remained an area for improvement. These results provide preliminary evidence of the technical feasibility of regulation-grounded conversational reporting. The study highlights opportunities to integrate regulatory knowledge into routine documentation, with future work needed to evaluate effects on report quality, safety awareness, and learning outcomes.

cs.HC↗