arXiv subjects

Wenhao Su

Publications and source records attributed to Wenhao Su.

3 recordsLinked to original sources

arXiv2026

SwarmX: Agentic Scheduling for Low-Latency Agentic Systems

Agentic AI applications compose multiple model calls and tool executions, creating new scheduling challenges for GPU-CPU clusters. Their inference time and model-call structure often depend on prompt semantics, making conventional scheduling approaches ineffective for low-latency serving. This paper presents SwarmX, a system that implements agentic scheduling for low-latency agentic applications. SwarmX uses scheduling-specific neural predictors to capture prompt, device, runtime, and target-model features; exposes distributional predictions to routers and scalers for tail-aware decisions; and provides mechanisms for predictor training and online adaptation. These predictors and mechanisms are integrated into a scheduler-agent framework that provides a common substrate for integration with existing scheduling and model-serving infrastructure. We evaluate SwarmX using production deployment (nearly one thousand GPUs and one million CPU cores) and controlled experiments on a 128-GPU testbed. Across multi-agent code generation, deep research, and multimodal agentic workflows, SwarmX reduces tail latency by up to 61.5% compared to state-of-the-art schedulers and sustains up to 2x the throughput of production schedulers under the same SLO.

cs.DC↗

arXiv2026

Modulating nonlinear optical responses in 3R-MoS$_2$ Fabry-P\'erot microcavities

Rhombohedrally stacked transition metal dichalcogenides such as 3R-MoS$_2$ offer an exceptional platform for nonlinear optics, naturally forming Fabry-P\'erot (FP) microcavities due to their giant dielectric contrast with the surrounding media. However, rigorously tracking the evolution of multiple harmonic fields within these unpatterned monolithic crystals remains a fundamental challenge. Here, we establish a self-consistent framework, spanning from linear broadband reflectance to second- and third-harmonic generation (SHG and THG), to systematically decode these nonlinear behaviors. Moving beyond conventional models, we demonstrate that the nonlinear emission is dictated by a delicate interplay among the intrinsic material absorption, the FP effects at the fundamental frequency, as well as those at the harmonic frequencies. When harmonic photons lie below the bandgap, weak absorption allows the nonlinear spectra to exhibit a complex modulation driven by the synergistic contribution of FP effects from both fundamental and harmonic waves. In stark contrast, severe intrinsic absorption of higher-energy photons heavily damps the FP effects of the harmonic fields, reducing the nonlinear response to an absorption-limited regime modulated almost exclusively by the FP effects at the fundamental frequency. By successfully decoupling these geometric and material contributions across different harmonic orders, our findings provide a precise design paradigm for engineering next-generation van der Waals photonic architectures.

physics.optics↗

arXiv2026

The Shadow Self: Intrinsic Value Misalignment in Large Language Model Agents

Large language model (LLM) agents with extended autonomy unlock new capabilities, but also introduce heightened challenges for LLM safety. In particular, an LLM agent may pursue objectives that deviate from human values and ethical norms, a risk known as value misalignment. Existing evaluations primarily focus on responses to explicit harmful input or robustness against system failure, while value misalignment in realistic, fully benign, and agentic settings remains largely underexplored. To fill this gap, we first formalize the Loss-of-Control risk and identify the previously underexamined Intrinsic Value Misalignment (Intrinsic VM). We then introduce IMPRESS (Intrinsic Value Misalignment Probes in REalistic Scenario Set), a scenario-driven framework for systematically assessing this risk. Following our framework, we construct benchmarks composed of realistic, fully benign, and contextualized scenarios, using a multi-stage LLM generation pipeline with rigorous quality control. We evaluate Intrinsic VM on 21 state-of-the-art LLM agents and find that it is a common and broadly observed safety risk across models. Moreover, the misalignment rates vary by motives, risk types, model scales, and architectures. While decoding strategies and hyperparameters exhibit only marginal influence, contextualization and framing mechanisms significantly shape misalignment behaviors. Finally, we conduct human verification to validate our automated judgments and assess existing mitigation strategies, such as safety prompting and guardrails, which show instability or limited effectiveness. We further demonstrate key use cases of IMPRESS across the AI Ecosystem. Our code and benchmark will be publicly released upon acceptance.

cs.CL↗