arXiv subjects

Jinhao Zhou

Publications and source records attributed to Jinhao Zhou.

4 recordsLinked to original sources

arXiv2026

Strategic Calibration in LLM Multi-Agent Systems: Audited Pivot Scoring for Truthful Elicitation

In large language model (LLM)-based multi-agent systems, reported confidence increasingly determines the allocation of scarce downstream attention, computation, and decision influence. This makes reporting strategic: an agent may overstate the relevance of its contribution even when its underlying belief is well calibrated. Existing calibration methods largely assume passive reporting and therefore do not address strategic incentives created when confidence determines downstream allocation. We formulate strategic calibration as a reporting-layer incentive problem and propose the Audited Pivot Scoring Mechanism (APSM). APSM offsets platform-controlled allocation benefits, uses inverse-propensity-weighted sparse audits, charges the Clarke pivot externality, and applies a strictly proper score. Under the stated model assumptions, truthful reporting is the unique expected-utility best response. We also characterize approximate execution and the audit-variance trade-off. Calculator-free experiments show that LLM agents inflate confidence under Top-$K$ incentives. With a shared agent-local utility interface, LLM agents reduce distortion under both Compensated Proper Scoring (CPS) and APSM. CPS achieves higher exact allocation match, whereas APSM yields lower Brier loss and numerically fewer cutoff crossings. Experiments further show that APSM enlarges the utility margin against allocation-changing deviations by $1.93\times$ and improves robustness under approximate execution. These results support incentive-aware elicitation as a distinct layer for reliable multi-agent coordination.

cs.GT↗

arXiv2026

When Messages Go Wrong: Understanding Communication Faults in LLM-Based Multi-Agent Systems

LLM-based multi-agent systems (LLM-MAS) use specialized agents that exchange messages to accomplish complex tasks. Their reliance on networked protocols, remote tools, and model endpoints exposes communication to message loss, delay, and context expiration. Because agents often exchange open-ended natural language, anomalies can alter task-relevant information while leaving messages processable, allowing semantic errors to propagate through subsequent interactions. Systematic studies of faults in inter-agent communication remain limited. This paper presents a large-scale empirical study of how communication faults manifest, propagate, and affect task outcomes in LLM-MAS. We construct a cross-layer taxonomy linking lower-stack disturbances to application-visible faults and develop a fault-injection and tracing pipeline. Our evaluation covers more than 29,000 executions across 100 tasks, five domains, three workflow topologies, and three model configurations. We find that 20.7% of failures with complete detection records occur without an immediate failure signal, reaching 76.2% for partial-content faults. Flat maintains 53.0% task success under communication faults, compared with 3.3% and 3.2% for Sequential and Hierarchical, respectively. This advantage primarily arises when an alternative information path preserves the evidence required by downstream agents. We also evaluate execution-side and evidence protection mechanisms and find that they recover different failure cases. Evidence recovery depends on whether trustworthy information remains accessible through an unaffected path. These observations provide empirical insights into communication-fault propagation, workflow resilience, and protection design for LLM-MAS.

cs.SE↗

arXiv2026

SparLeak: Privacy Leakage from Sparse Attention in LLM Inference on Shared GPUs

Sparse attention is widely used to accelerate long-context inference in modern large language models (LLMs), but its input-dependent execution behavior introduces previously unexplored privacy risks. We identify a new GPU micro-architectural side channel, termed Sparsity-Induced Memory Access (SIMA), which arises from secret-dependent key-value cache access patterns induced by sparse attention. Based on this observation, we present SparLeak, a phase-aware side-channel attack that extracts SIMA traces during LLM inference and enables two practical privacy extractions: query attribute inference from prefill-phase traces and autoregressive response reconstruction from decoding-phase traces. By reconstructing approximate token-level sparsity profiles from page-level observations and applying profiling-based learning, SparLeak accurately recovers sensitive information, including user-query attributes and private LLM response content. Extensive evaluation across three LLM architectures, three sparse attention mechanisms, and three privacy-sensitive datasets shows that SparLeak achieves average attack success rates of 90.9% for attribute inference and 87.3% for response reconstruction under real-world LLM serving settings, highlighting the significance to account for SIMA leakage when deploying sparse-attention-based LLM systems. We provide anonymized SIMA traces, trained attack models, evaluation scripts, and documentation as artifacts at https://anonymous.4open.science/r/Janus_artifacts/.

cs.LG↗

arXiv2025

Flash-VL 2B: Optimizing Vision-Language Model Performance for Ultra-Low Latency and High Throughput

In this paper, we introduce Flash-VL 2B, a novel approach to optimizing Vision-Language Models (VLMs) for real-time applications, targeting ultra-low latency and high throughput without sacrificing accuracy. Leveraging advanced architectural enhancements and efficient computational strategies, Flash-VL 2B is designed to maximize throughput by reducing processing time while maintaining competitive performance across multiple vision-language benchmarks. Our approach includes tailored architectural choices, token compression mechanisms, data curation, training schemes, and a novel image processing technique called implicit semantic stitching that effectively balances computational load and model performance. Through extensive evaluations on 11 standard VLM benchmarks, we demonstrate that Flash-VL 2B achieves state-of-the-art results in both speed and accuracy, making it a promising solution for deployment in resource-constrained environments and large-scale real-time applications.

cs.CV↗