arXiv subjects

Shizhen Zhao

Publications and source records attributed to Shizhen Zhao.

At least 19 recordsLinked to original sources

arXiv2026

Breaking the Duplex Barrier: Lane-Granularity OCS Scheduling for LLM Training

Optical circuit switch (OCS) can reconfigure physical connectivity to match the predictable communication schedules of large language model (LLM) training. Although each OCS light path is physically simplex, existing demand-aware OCS schedulers allocate capacity in duplex-port pairs, forcing equal bandwidth in both directions and stranding capacity under asymmetric node-pair traffic. This paper present LACE, the first offline OCS schedule compiler that independently allocates transmit (TX) and receive (RX) lanes for LLM training. Without changing the selected collective algorithms, operation order, or rank placement, LACE reconstructs directed node-level demand, jointly determines which consecutive operations share a configuration and how many simplex circuits serve each direction, and realizes these allocations as physical lane bindings and optical paths under per-node lane-inventory and multi-OCS fabric constraints. Software acknowledgments carry feedback over independently provisioned return paths, while coordinated link configuration and recovery verify each configuration before communication resumes. On a separate three-server testbed using fixed topologies and matched per-port rate limits, LACE's asymmetric connectivity achieves $1.80\times$ speedup for communication replay and $1.27\times$ for GPT-2 training over a symmetric-topology baseline. At larger scale, simulations of LLaMA-3.1 70B and 405B schedules with sixteen 400-Gb/s ports per server show that LACE achieves $1.21$--$2.04\times$ communication speedup over the latest duplex OCS scheduler.

cs.NI↗

arXiv2026

Dynamic Important Example Mining for Reinforcement Finetuning

Reinforcement fine-tuning (RFT) is increasingly used to strengthen the reasoning abilities of large models, yet its effectiveness is bound by how training data are selected and used. Most data-centric RFT methods rely on static or heuristic sample selection, implicitly assuming a sample's value is fixed over training. This overlooks the non-stationary dynamics of policy learning and can lead to suboptimal updates. We propose Dynamic Important Example Mining (DIEM), a principled and fully automated framework that makes data utilization adaptive throughout RFT. DIEM integrates two components into each optimization step: (i) a gradient-alignment importance estimator that efficiently approximates each sample's marginal contribution to policy improvement; and (ii) a constrained batch reweighting scheme that maximizes aggregate utility while preserving the update's gradient magnitude to stabilize optimization. Across several reasoning benchmarks, DIEM consistently outperforms strong static and dynamic baselines. The code will be released via https://github.com/hrtan/DIEM.

cs.AI↗

arXiv2026

Janus: An Algorithm-Evaluator Co-Evolution Framework for LLM-Driven Discovery under Expensive Evaluation Budgets

LLM-driven program discovery relies on rapid evaluator feedback, but many scientific and engineering tasks require high-fidelity simulations, hardware execution, or physical experiments, making each evaluation expensive. Cheap surrogate evaluators can reduce this cost, yet fixed surrogates are vulnerable to search-induced distribution shift and are difficult to fit reliably from sparse, search-biased labels. We introduce Janus, a framework that uses LLMs to co-evolve target programs and executable proxy evaluators. To address label scarcity, Janus leverages domain knowledge encoded in LLMs to generate task-specific evaluator programs and calibrates them using real outcomes. To mitigate distribution shift, Janus evolves evaluators alongside target programs, selects them using a promotion-aligned objective, and maintains region-conditioned portfolios with online credit updates. Because proxy predictions remain fallible, Janus uses them only to prioritize candidates and requires real validation before candidates can enter the target-program population or update the incumbent. Across five scientific and engineering design tasks, Janus achieves a larger area under the best-so-far improvement curve over the real-evaluation budget and higher final performance than a matched baseline that evolves only target programs. On average, Janus reaches 99/% of the baseline's final improvement with 59.1/% fewer real evaluations. Evolved proxy evaluators also rank promising candidates more accurately than their seed versions. Together, these results extend evaluator-guided LLM discovery from tasks with cheap, scalable feedback to scientific domains where trustworthy evaluation is scarce and expensive.

cs.AI↗

arXiv2026

Geometry-Instructed Video Editing

Object-level geometric edits, including translating, rotating, scaling, duplicating, or removing an object, are routine operations in digital content creation (DCC) workflows, yet they remain unreliable in generative video editing. The key challenge lies in specifying the target object's 3D state change unambiguously across viewpoint and time, while consistently updating geometry-dependent secondary effects such as shadows and reflections. We introduce GIVE, a geometry-instructed video editing framework that represents edits through a unified object-state formulation. Two video-aligned geometry streams describe the target object before and after editing: a depth-box encoding coarse 3D placement and extent, and an orientation-box providing an appearance-agnostic orientation cue. Together, these streams provide a compact pre/post geometric specification for object-state transitions. To provide paired supervision for learning these edits, we build a scalable graphics-engine pipeline that executes object-level edit programs and renders controlled before/after pairs, isolating the intended geometric edit while keeping secondary effects consistent with the transformation. Experimental results demonstrate that GIVE produces faithful geometric edits with temporal coherence and consistent secondary effects across operators in a unified framework, and shows promising transfer to in-the-wild videos. Project page: https://geometry-instructed-video-editing.github.io/give/

cs.CV↗

arXiv2026

LiveR: Fine-Grained Elasticity via Live Reconfiguration for Model Training

To reduce user costs and maximize cluster utilization, large model training increasingly leverages volatile but inexpensive GPU capacity, such as spot instances and reclaimable resources in shared clusters. Yet, capitalizing on these economic benefits requires jobs to adapt within the short warning windows that many such environments provide. Existing elastic training systems still treat reconfiguration as stop-and-restart: they externalize distributed state through checkpoints, rebuild the distributed runtime on a new topology, and restart training, turning each resize event into a storage-heavy recovery procedure that incurs substantial downtime from checkpoint I/O, process restart, CUDA initialization, and communicator setup. We present LiveR, a live reconfiguration runtime for elastic LLM training that replaces storage-backed restart with a live, bounded-memory handoff between mixed-parallel training worlds. While the current world continues training, LiveR asynchronously prepares the target world, bootstraps newly added workers in isolation to keep heavyweight initialization off the critical path, and streams model state directly over high-bandwidth interconnects while reshaping it online across tensor, pipeline, and data parallel dimensions. Once the target world is ready, LiveR performs a lightweight commit that switches training to the new configuration without stop-and-restart on the live path. We implement LiveR atop Megatron-LM and PyTorch and evaluate it end-to-end on a multi-node GPU cluster. Across diverse reconfiguration scenarios, LiveR reduces downtime from minutes to seconds, accelerates reconfiguration by 14$\times$-23$\times$ over checkpoint/restart baselines, incurs minimal steady-state overhead, and sustains up to 99% training goodput under volatile-resource conditions, making volatile low-cost GPU capacity far more practical for LLM training.

cs.DC↗

arXiv2026

NeuroRisk: Physics-Informed Neural Optimization for Risk-Aware Traffic Engineering

In production Wide-Area Networks (WANs), correlated failures dominate availability losses, forcing operators to reserve large safety margins that leave substantial capacity underutilized. Achieving high utilization under strict availability targets therefore requires risk-aware Traffic Engineering (TE) over dozens to hundreds of probabilistic failure scenarios-yet solving this problem at operational timescales remains elusive. We demonstrate that existing risk-aware formulations can be unified under an embedded Sort-and-Select structure, exposing a fundamental trade-off between expressiveness and tractability: classical optimizers either restrict scenario selection for efficiency or incur prohibitive decomposition costs. While deep learning appears promising, prior Deep TE methods mainly target maximum link utilization and rely on scaling-based feasibility, which fundamentally breaks under explicit capacity constraints and scenario-dependent risk. We present NeuroRisk, a physics-informed deep unrolled optimizer that exploits the structure of Sort-and-Select. NeuroRisk enforces feasibility via gated edge-local reservations and represents scenario sets through permutation-invariant, gradient-aligned cues. Evaluations on production-style WANs show that NeuroRisk achieves small optimality gaps relative to the solver with orders of magnitude speedup $(10^2- 10^5 \times)$ on risk objectives, while outperforming neural baselines on nominal throughput.

cs.NI↗

arXiv2026

Libra: Accelerating Socket I/O via Programmable Selective Data Copying

Layer-7 (L7) proxies are critical to modern cloud-native systems, yet their performance is increasingly bottlenecked by copying entire payloads across the kernel-user boundary. Existing approaches reduce this overhead but typically sacrifice compatibility with unmodified POSIX applications, introduce new APIs, or require specialized environments. We show that, under conventional OS abstractions, fully eliminating kernel-user copies while preserving standard socket semantics for unmodified proxies is fundamentally impossible. This leads to a practical insight: in common L7 workloads, proxies inspect only small metadata (e.g., HTTP headers) for routing, while forwarding the bulk payload unchanged. Based on this insight, we present Libra, an OS-level selective-copy framework that copies only metadata to the user space and retains the bulk payload in the kernel for forwarding, reducing data movement without breaking compatibility. Libra uses eBPF to identify protocol-specific metadata boundaries and coordinate selective copy and payload reuse across receive and transmit paths, all without modifying the socket API. Implemented in Linux and evaluated with unmodified Nginx and HAProxy, Libra improves plaintext throughput by up to 4.2x and reduces P99 tail latency by over 90%. With hardware-offloaded kTLS, it boosts encrypted throughput by 2.0x and cuts tail latency by 65%.

cs.NI↗

arXiv2026

FEPLB: Exploiting Copy Engines for Nearly Free MoE Load Balancing in Distributed Training

Fine-grained, per-micro-batch load balancing is essential for efficient Mixture-of-Experts (MoE) training, yet every prior dynamic scheduling scheme pays for it with extra communication that is hard to hide. Especially on modern bulk-transfer backends such as DeepEP. We make a simple but consequential observation: on the NVIDIA Hopper architecture the NVLink Copy Engine can move data between intra-node GPUs without consuming any SM cycles, effectively providing a nearly free communication channel that runs in parallel with compute kernels. FEPLB turns this idle hardware into a new parallel dimension for MoE load rebalancing. Its Two-Phase Dispatch first routes tokens across nodes via the standard EP backend, then redistributes dynamic-expert tokens and weights within the NVLink domain through the Copy Engine at nearly zero cost, while a lightweight CPU scheduler runs concurrently with static expert computation. Because FEPLB uses only Copy Engine and CPU that are orthogonal to those consumed by EP and PP, it coexists with existing parallel strategies without reconfiguration. On GLM-5's MoE layers (128 experts, no auxiliary loss, up to 16 H100 GPUs), FEPLB reduces the token straggler by 51-70% and the GEMM straggler by 50-68% with no measurable EP communication overhead. Its advantage grows with the EP degree: at EP=8, it achieves 2x lower token straggler than FasterMoE.

cs.DC↗

arXiv2026

Geminet: Learning the Duality-based Iterative Process for Lightweight Traffic Engineering in Changing Topologies

Recently, researchers have explored ML-based Traffic Engineering (TE), leveraging neural networks to solve TE problems traditionally addressed by optimization. However, existing ML-based TE schemes remain impractical: they either fail to handle topology changes or suffer from poor scalability due to excessive computational and memory overhead. To overcome these limitations, we propose Geminet, a lightweight and scalable ML-based TE framework that can handle changing topologies. Geminet is built upon two key insights: (i) a methodology that decouples neural networks from topology by learning an iterative gradient-descent-based adjustment process, as the update rule of gradient descent is topology-agnostic, relying only on a few gradient-related quantities; (ii) shifting optimization from path-level routing weights to edge-level dual variables, reducing memory consumption by leveraging the fact that edges are far fewer than paths. Evaluations on WAN and data center datasets show that Geminet significantly improves scalability. Its neural network size is only 0.04% to 7% of existing schemes, while handling topology variations as effectively as HARP, a state-of-the-art ML-based TE approach, without performance degradation. When trained on large-scale topologies, Geminet consumes under 10 GiB of memory, more than eight times less than the 80-plus GiB required by HARP, while achieving 5.45 times faster convergence speed, demonstrating its potential for large-scale deployment.

cs.NI↗

arXiv2026

Leaf-centric Logical Topology Design for OCS-based GPU Clusters

Recent years have witnessed the growing deployment of optical circuit switches (OCS) in commercial GPU clusters (e.g., Google A3 GPU cluster) optimized for machine learning (ML) workloads. Such clusters adopt a three-tier leaf-spine-OCS topology, servers attach to leaf-layer electronic packet switches (EPSes); these leaf switches aggregate into spine-layer EPSes to form a Pod; and multiple Pods are interconnected via core-layer OCSes. Unlike EPSes, OCSes only support circuit-based paths between directly connected spine switches, potentially inducing a phenomenon termed routing polarization, which refers to the scenario where the bandwidth requirements between specific pairs of Pods are unevenly fulfilled through links among different spine switches. The resulting imbalance induces traffic contention and bottlenecks on specific leaf-to-spine links, ultimately reducing ML training throughput. To mitigate this issue, we introduce a leaf-centric paradigm to ensure traffic originating from the same leaf switch is evenly distributed across multiple spine switches with balanced loads. Through rigorous theoretical analysis, we establish a sufficient condition for avoiding routing polarization and propose a corresponding logical topology design algorithm with polynomial-time complexity. Large-scale simulations validate up to 19.27% throughput improvement and a 99.16% reduction in logical topology computation overhead compared to Mixed Integer Programming (MIP)-based methods.

cs.NI↗

arXiv2026

SHIFT: Exploring the Boundary of RDMA Network Fault Tolerance

Under gang scheduling for large-scale distributed large language model (LLM) training, a single network anomaly can stall or abort an entire job. Current network fault tolerance mechanisms typically adopt a ``fallback and bypass'' approach within the switching fabric and at the access layer, tolerating in-network and access-layer failures. We explore whether RDMA fault tolerance can be extended to the cross-NIC level by failing over traffic to intra-host backup NICs. For the first time, we prove a fundamental Trilemma: it is impossible to have Cross-NIC RDMA failover that simultaneously preserves Exactly-Once Execution, Receiver-NIC Opacity, and a Zero-Copy datapath. Fortunately, we observe that dominant training frameworks (e.g., NCCL) rely on idempotent bulk transfers that tolerate relaxed memory ordering, as long as notification ordering is preserved. Leveraging this insight, we present SHIFT, a user-space RDMA layer that provides cross-NIC fault tolerance while preserving correct memory semantics. We implement SHIFT in \texttt{rdma-core} and evaluate it with PyTorch distributed training. Results show that SHIFT incurs negligible overhead during normal operation and successfully masks fatal NIC failures and link anomalies, allowing training to continue without costly restarts.

cs.NI↗

arXiv2026

FastReChain: A Novel Bidirectional Model-Based Algorithm for Topology Engineering of OCS-Based Clusters

Optical Circuit Switching (OCS) technology is increasingly being adopted in data centers due to its advantages of low power consumption and low technology refresh costs. Unlike electrical packet switches, OCS provides programmable bandwidth for directly connected devices by configuring the mapping relationships of internal ports. Thus, how to calculate these internal port mapping relationships, i.e., Topology Engineering (ToE), is one of the key designs of OCS-based clusters. Current deployments usually design ToE algorithms by solving Integer Linear Programming (ILP) models, with the aim of minimizing modifications to links occupied by running tasks as much as possible. However, ILP-based ToE algorithms may incur excessive runtime overhead in large-scale clusters. Some existing ToE algorithms convert the ILP model into a Minimum-Cost Flow model through greedy construction, yet such greedy strategies may increase the number of affected links during the OCS reconfiguration process. To solve the aforementioned problems, we propose a novel bidirectional modeling approach, along with a corresponding FastReChain algorithm in this paper. We verify the superiority of this algorithm through simulation experiments based on real-trace data.

cs.NI↗

arXiv2026

SAIR: Cost-Efficient Multi-Stage ML Pipeline Autoscaling via In-Context Reinforcement Learning

Multi-stage ML inference pipelines are difficult to autoscale due to heterogeneous resources, cross-stage coupling, and dynamic bottleneck migration. We present SAIR, an autoscaling framework that uses an LLM as an in-context reinforcement learning controller, improving its policy online from reward-labeled interaction histories without gradient updates. SAIR combines Pareto-dominance reward shaping with a provable separation margin, surprisal-guided experience retrieval for context efficiency, and fine-grained GPU rate control via user-space CUDA interception. We provide regret analysis decomposing error into retrieval coverage and LLM selection components. On four ML serving pipelines under three workload patterns, SAIR achieves the best or tied-best P99 latency and effective resource cost among deployed baselines, improving P99 by up to 50% and reducing effective cost by up to 97% (under GPU rate-control assumptions), with 86% bottleneck detection accuracy and no offline training.

cs.LG↗

arXiv2025

ATRO: A Fast Algorithm for Topology Engineering of Reconfigurable Datacenter Networks

Reconfigurable data center networks (DCNs) enhance traditional architectures with optical circuit switches (OCSs), enabling dynamic reconfiguration of inter-pod links, i.e., the logical topology. Optimizing this topology is crucial for adapting to traffic dynamics but is challenging due to its combinatorial nature. The complexity increases further when demands can be distributed across multiple paths, requiring joint optimization of topology and routing. We propose Alternating Topology and Routing Optimization (ATRO), a unified framework that supports both one-hop topology optimization (where traffic is routed via direct paths) and multi-hop joint optimization (where routing is also optimized). Although these settings differ in constraints, both are combinatorially hard and challenge solver-based methods. ATRO addresses both cases efficiently: in the one-hop case, it guarantees the global optimum via an accelerated binary search; in the multi-hop case, it alternates between topology and routing updates, with routing steps optionally accelerated by existing traffic engineering (TE) methods. ATRO supports warm-starting and improves solution quality monotonically across iterations. ATRO remains competitive even when paired with solver-free TE methods, forming a fully solver-free optimization pipeline that still outperforms prior approaches in runtime and maximum link utilization across diverse workloads.

cs.NI↗

arXiv2025

ASSIST-3D: Adapted Scene Synthesis for Class-Agnostic 3D Instance Segmentation

Class-agnostic 3D instance segmentation tackles the challenging task of segmenting all object instances, including previously unseen ones, without semantic class reliance. Current methods struggle with generalization due to the scarce annotated 3D scene data or noisy 2D segmentations. While synthetic data generation offers a promising solution, existing 3D scene synthesis methods fail to simultaneously satisfy geometry diversity, context complexity, and layout reasonability, each essential for this task. To address these needs, we propose an Adapted 3D Scene Synthesis pipeline for class-agnostic 3D Instance SegmenTation, termed as ASSIST-3D, to synthesize proper data for model generalization enhancement. Specifically, ASSIST-3D features three key innovations, including 1) Heterogeneous Object Selection from extensive 3D CAD asset collections, incorporating randomness in object sampling to maximize geometric and contextual diversity; 2) Scene Layout Generation through LLM-guided spatial reasoning combined with depth-first search for reasonable object placements; and 3) Realistic Point Cloud Construction via multi-view RGB-D image rendering and fusion from the synthetic scenes, closely mimicking real-world sensor data acquisition. Experiments on ScanNetV2, ScanNet++, and S3DIS benchmarks demonstrate that models trained with ASSIST-3D-generated data significantly outperform existing methods. Further comparisons underscore the superiority of our purpose-built pipeline over existing 3D scene synthesis approaches.

cs.CV↗

arXiv2025

Aligning Effective Tokens with Video Anomaly in Large Language Models

Understanding abnormal events in videos is a vital and challenging task that has garnered significant attention in a wide range of applications. Although current video understanding Multi-modal Large Language Models (MLLMs) are capable of analyzing general videos, they often struggle to handle anomalies due to the spatial and temporal sparsity of abnormal events, where the redundant information always leads to suboptimal outcomes. To address these challenges, exploiting the representation and generalization capabilities of Vison Language Models (VLMs) and Large Language Models (LLMs), we propose VA-GPT, a novel MLLM designed for summarizing and localizing abnormal events in various videos. Our approach efficiently aligns effective tokens between visual encoders and LLMs through two key proposed modules: Spatial Effective Token Selection (SETS) and Temporal Effective Token Generation (TETG). These modules enable our model to effectively capture and analyze both spatial and temporal information associated with abnormal events, resulting in more accurate responses and interactions. Furthermore, we construct an instruction-following dataset specifically for fine-tuning video-anomaly-aware MLLMs, and introduce a cross-domain evaluation benchmark based on XD-Violence dataset. Our proposed method outperforms existing state-of-the-art methods on various benchmarks.

cs.CV↗

arXiv2025

Equipping Vision Foundation Model with Mixture of Experts for Out-of-Distribution Detection

Pre-trained vision foundation models have transformed many computer vision tasks. Despite their strong ability to learn discriminative and generalizable features crucial for out-of-distribution (OOD) detection, their impact on this task remains underexplored. Motivated by this gap, we systematically investigate representative vision foundation models for OOD detection. Our findings reveal that a pre-trained DINOv2 model, even without fine-tuning on in-domain (ID) data, naturally provides a highly discriminative feature space for OOD detection, achieving performance comparable to existing state-of-the-art methods without requiring complex designs. Beyond this, we explore how fine-tuning foundation models on in-domain (ID) data can enhance OOD detection. However, we observe that the performance of vision foundation models remains unsatisfactory in scenarios with a large semantic space. This is due to the increased complexity of decision boundaries as the number of categories grows, which complicates the optimization process. To mitigate this, we propose the Mixture of Feature Experts (MoFE) module, which partitions features into subspaces, effectively capturing complex data distributions and refining decision boundaries. Further, we introduce a Dynamic-$β$ Mixup strategy, which samples interpolation weights from a dynamic beta distribution. This adapts to varying levels of learning difficulty across categories, improving feature learning for more challenging categories. Extensive experiments demonstrate the effectiveness of our approach, significantly outperforming baseline methods.

cs.CV↗

arXiv2025

A Highly Scalable LLM Clusters with Optical Interconnect

Recent years have witnessed the adoption of optical circuit switch (OCS) technology. How to design the physical topology, defined by the physical wiring between electrical switching equipments and the OCS, is fundamental to designing efficient OCS-based clusters. We identify three features to evaluate the quality of a physical topology design: logical topology compatibility, cluster scalability, and topology engineering polynomial-solvability. However, none of existing physical topologies has achieved these three features simultaneously. This paper explores designing an optimal physical topology that simultaneously maximizes all. We begin by analyzing the importance of these features in OCS-based cluster and examine the limitations of current designs. Leveraging a proposed \emph{Symmetric Integer Matrix Decomposition Theorem}, we outline a general approach for designing optimal physical topologies and introduce \textbf{Cross Wiring} as a concrete instantiation. The feasibility and advantages of Cross Wiring are verified through a 128-NPU testbed and large-scale real-trace-based simulations.

cs.NI↗