arXiv subjects

Sheng Guan

Publications and source records attributed to Sheng Guan.

14 recordsLinked to original sources

arXiv2026

LiveServe: Interaction-Aware Serving for Real-Time Omni-Modal LLMs

Realtime omni-modal LMs support speech-centric conversations where users stream inputs, hear generated audio, and interrupt freely. Existing Omni-LM serving systems still rely on throughput-oriented LLM scheduling and LRU KV offloading. These policies ignore audio playback and multi-turn reuse: they may generate tokens far beyond what users hear, wasting work after barge-in, and evict KV state needed in the next turn. LiveServe is an interaction-aware serving system for realtime Omni-LM interaction. It exposes playback progress, speech activity, and barge-in events to the serving pipeline. The scheduler prioritizes first-audio and near-underrun sessions while limiting generation beyond the playback frontier. The KV manager uses next-use-aware eviction and preloads likely-needed KV during user speech to hide reload latency. On vLLM-Omni, LiveServe improves realtime serving across two Omni-LMs and mixed workloads. It lowers P90 audio TTFP by $1.55\times$ on average and up to $2.21\times$, while improving completed-request throughput by $1.15\times$ on average and up to $1.56\times$, and moves most KV reload work off the next-turn critical path.

cs.DC↗

arXiv2026

RecMem: Recurrence-based Memory Consolidation for Efficient and Effective Long-Running LLM Agents

Memory systems often organize user-agent interactions as retrievable external memory and are crucial for long-running agents by overcoming the limited context windows of LLMs. However, existing memory systems invoke LLMs to process every incoming interaction for memory extraction, and such an eager memory consolidation scheme leads to substantial token consumption. To tackle this problem, we propose RecMem by rethinking when memory consolidation should be conducted. RecMem stores incoming interactions in a subconscious memory layer and encode them using lightweight embedding models for retrieval. LLMs are only invoked to extract episodic and semantic memory when sustained recurrence are observed for semantically similar interactions. Such recurrence-based consolidation works because these interactions correspond to a semantic cluster with rich information and thus are worth extraction and summarization. To improve accuracy, RecMem also incorporates a semantic refinement mechanism that recovers the fine-grained facts omitted by memory extraction. Experiments show that RecMem reduces the memory construction token cost of three SOTA memory systems by up to 87% while exceeding their accuracy.

cs.CL↗

arXiv2026

Scalable Prompt Routing via Fine-Grained Latent Task Discovery

Prompt routing dynamically selects the most appropriate large language model from a pool of candidates for each query, optimizing performance while managing costs. As model pools scale to include dozens of frontier models with narrow performance gaps, existing approaches face significant challenges: manually defined task taxonomies cannot capture fine-grained capability distinctions, while monolithic routers struggle to differentiate subtle differences across diverse tasks. We propose a two-stage routing architecture that addresses these limitations through automated fine-grained task discovery and task-aware quality estimation. Our first stage employs graph-based clustering to discover latent task types and trains a classifier to assign prompts to discovered tasks. The second stage uses a mixture-of-experts architecture with task-specific prediction heads for specialized quality estimates. At inference, we aggregate predictions from both stages to balance task-level stability with prompt-specific adaptability. Evaluated on 10 benchmarks with 11 frontier models, our method consistently outperforms existing baselines and surpasses the strongest individual model while incurring less than half its cost.

cs.CL↗

arXiv2026

Consensus is Not Verification: Why Crowd Wisdom Strategies Fail for LLM Truthfulness

Pass@k and other methods of scaling inference compute can improve language model performance in domains with external verifiers, including mathematics and code, where incorrect candidates can be filtered reliably. This raises a natural question: can we similarly scale compute to elicit gains in truthfulness for domains without convenient verification? We show that across five benchmarks and models, surprisingly, it cannot. Even at 25x the inference cost of naive sampling, polling-style aggregation yields no consistent accuracy gains over single-sample baselines and often amplifies shared misconceptions. We find that under uncertainty, models are better at predicting what other models will say within model ensembles than at identifying what is true, revealing a separation between social prediction and truth verification. Across models and benchmarks, aggregation fails to provide a robust truth signal because language model errors are strongly correlated. The source of correlation goes beyond any individual benchmark: we show that even when conditioned on out of distribution random strings and asked to produce pseudo-random outputs, different models produce correlated outputs. Confidence-based weighting provides no benefit because self-reported confidence fails to reliably distinguish correct from incorrect answers. These results delineate a boundary for inference-time scaling: in verified domains, additional samples provide more candidates for a verifier to filter; in unverified domains, additional samples merely reinforce shared misconceptions.

cs.LG↗

arXiv2026

Train Less, Learn More: Adaptive Efficient Rollout Optimization for Group-Based Reinforcement Learning

Reinforcement learning (RL) plays a central role in large language model (LLM) post-training. Among existing approaches, Group Relative Policy Optimization (GRPO) is widely used, especially for RL with verifiable rewards (RLVR) fine-tuning. In GRPO, each query prompts the LLM to generate a group of rollouts with a fixed group size $N$. When all rollouts in a group share the same outcome, either all correct or all incorrect, the group-normalized advantages become zero, yielding no gradient signal and wasting fine-tuning compute. We introduce Adaptive Efficient Rollout Optimization (AERO), an enhancement of GRPO. AERO uses an adaptive rollout strategy, applies selective rejection to strategically prune rollouts, and maintains a Bayesian posterior to prevent zero-advantage dead zones. Across three model configurations (Qwen2.5-Math-1.5B, Qwen2.5-7B, and Qwen2.5-7B-Instruct), AERO improves compute efficiency without sacrificing performance. Under the same total rollout budget, AERO reduces total training compute by about 48% while shortening wall-clock time per step by about 45% on average. Despite the substantial reduction in compute, AERO matches or improves Pass@8 and Avg@8 over GRPO, demonstrating a practical, scalable, and compute-efficient strategy for RL-based LLM alignment.

cs.LG↗

arXiv2025

MarkDiffusion: An Open-Source Toolkit for Generative Watermarking of Latent Diffusion Models

We introduce MarkDiffusion, an open-source Python toolkit for generative watermarking of latent diffusion models. It comprises three key components: a unified implementation framework for streamlined watermarking algorithm integrations and user-friendly interfaces; a mechanism visualization suite that intuitively showcases added and extracted watermark patterns to aid public understanding; and a comprehensive evaluation module offering standard implementations of 24 tools across three essential aspects - detectability, robustness, and output quality - plus 8 automated evaluation pipelines. Through MarkDiffusion, we seek to assist researchers, enhance public awareness and engagement in generative watermarking, and promote consensus while advancing research and applications.

cs.CR↗

arXiv2025

IPR: Intelligent Prompt Routing with User-Controlled Quality-Cost Trade-offs

Routing incoming queries to the most cost-effective LLM while maintaining response quality poses a fundamental challenge in optimizing performance-cost trade-offs for large-scale commercial systems. We present IPR\, -- \,a quality-constrained \textbf{I}ntelligent \textbf{P}rompt \textbf{R}outing framework that dynamically selects optimal models based on predicted response quality and user-specified tolerance levels. IPR introduces three key innovations: (1) a modular architecture with lightweight quality estimators trained on 1.5M prompts annotated with calibrated quality scores, enabling fine-grained quality prediction across model families; (2) a user-controlled routing mechanism with tolerance parameter $\tau \in [0,1]$ that provides explicit control over quality-cost trade-offs; and (3) an extensible design using frozen encoders with model-specific adapters, reducing new model integration from days to hours. To rigorously train and evaluate IPR, we curate an industrial-level dataset IPRBench\footnote{IPRBench will be released upon legal approval.}, a comprehensive benchmark containing 1.5 million examples with response quality annotations across 11 LLM candidates. Deployed on a major cloud platform, IPR achieves 43.9\% cost reduction while maintaining quality parity with the strongest model in the Claude family and processes requests with sub-150ms latency. The deployed system and additional product details are publicly available at https://aws.amazon.com/bedrock/intelligent-prompt-routing/

cs.LG↗

arXiv2025

Omni-SafetyBench: A Benchmark for Safety Evaluation of Audio-Visual Large Language Models

Omni-modal Large Language Models (OLLMs) that integrate visual, auditory, and textual processing face severe safety risks. They exhibit fragile defenses against audio-visual joint harmful inputs and demonstrate inconsistent safety performance across different modalities, enabling simple modality-switching jailbreaks. However, existing safety benchmarks fail to comprehensively assess these risks due to the absence of audio-visual joint samples, limited modality coverage, and lack of parallel test cases for cross-modal consistency evaluation. To address these gaps, we introduce Omni-SafetyBench, the first comprehensive parallel benchmark for OLLM safety evaluation, featuring 23,328 test instances across 24 modality variations derived from 972 seed samples. Recognizing that complex inputs pose comprehension challenges and that cross-modal consistency is critical for OLLM safety, we propose tailored metrics: a Safety-score based on Conditional Attack Success Rate (C-ASR) and Conditional Refusal Rate (C-RR), and a Cross-Modal Safety Consistency score (CMSC-score). Evaluating 11 state-of-the-art OLLMs reveals severe vulnerabilities: only 3 models exceed 0.6 in both metrics, with safety degrading sharply for audio-visual inputs. Furthermore, evaluation of existing safety alignment methods on Omni-SafetyBench identifies fundamental challenges in OLLM safety alignment, highlighting urgent needs for enhanced research in this domain.

cs.CL↗

arXiv2025

Black-Box Visual Prompt Engineering for Mitigating Object Hallucination in Large Vision Language Models

Large Vision Language Models (LVLMs) often suffer from object hallucination, which undermines their reliability. Surprisingly, we find that simple object-based visual prompting -- overlaying visual cues (e.g., bounding box, circle) on images -- can significantly mitigate such hallucination; however, different visual prompts (VPs) vary in effectiveness. To address this, we propose Black-Box Visual Prompt Engineering (BBVPE), a framework to identify optimal VPs that enhance LVLM responses without needing access to model internals. Our approach employs a pool of candidate VPs and trains a router model to dynamically select the most effective VP for a given input image. This black-box approach is model-agnostic, making it applicable to both open-source and proprietary LVLMs. Evaluations on benchmarks such as POPE and CHAIR demonstrate that BBVPE effectively reduces object hallucination.

cs.CV↗

arXiv2025

A Systematic Survey of Automatic Prompt Optimization Techniques

Since the advent of large language models (LLMs), prompt engineering has been a crucial step for eliciting desired responses for various Natural Language Processing (NLP) tasks. However, prompt engineering remains an impediment for end users due to rapid advances in models, tasks, and associated best practices. To mitigate this, Automatic Prompt Optimization (APO) techniques have recently emerged that use various automated techniques to help improve the performance of LLMs on various tasks. In this paper, we present a comprehensive survey summarizing the current progress and remaining challenges in this field. We provide a formal definition of APO, a 5-part unifying framework, and then proceed to rigorously categorize all relevant works based on their salient features therein. We hope to spur further research guided by our framework.

cs.CL↗

arXiv2025

Refining Positive and Toxic Samples for Dual Safety Self-Alignment of LLMs with Minimal Human Interventions

Recent AI agents, such as ChatGPT and LLaMA, primarily rely on instruction tuning and reinforcement learning to calibrate the output of large language models (LLMs) with human intentions, ensuring the outputs are harmless and helpful. Existing methods heavily depend on the manual annotation of high-quality positive samples, while contending with issues such as noisy labels and minimal distinctions between preferred and dispreferred response data. However, readily available toxic samples with clear safety distinctions are often filtered out, removing valuable negative references that could aid LLMs in safety alignment. In response, we propose PT-ALIGN, a novel safety self-alignment approach that minimizes human supervision by automatically refining positive and toxic samples and performing fine-grained dual instruction tuning. Positive samples are harmless responses, while toxic samples deliberately contain extremely harmful content, serving as a new supervisory signals. Specifically, we utilize LLM itself to iteratively generate and refine training instances by only exploring fewer than 50 human annotations. We then employ two losses, i.e., maximum likelihood estimation (MLE) and fine-grained unlikelihood training (UT), to jointly learn to enhance the LLM's safety. The MLE loss encourages an LLM to maximize the generation of harmless content based on positive samples. Conversely, the fine-grained UT loss guides the LLM to minimize the output of harmful words based on negative samples at the token-level, thereby guiding the model to decouple safety from effectiveness, directing it toward safer fine-tuning objectives, and increasing the likelihood of generating helpful and reliable content. Experiments on 9 popular open-source LLMs demonstrate the effectiveness of our PT-ALIGN for safety alignment, while maintaining comparable levels of helpfulness and usefulness.

cs.CL↗

arXiv2024

Can Watermarked LLMs be Identified by Users via Crafted Prompts?

Text watermarking for Large Language Models (LLMs) has made significant progress in detecting LLM outputs and preventing misuse. Current watermarking techniques offer high detectability, minimal impact on text quality, and robustness to text editing. However, current researches lack investigation into the imperceptibility of watermarking techniques in LLM services. This is crucial as LLM providers may not want to disclose the presence of watermarks in real-world scenarios, as it could reduce user willingness to use the service and make watermarks more vulnerable to attacks. This work is the first to investigate the imperceptibility of watermarked LLMs. We design an identification algorithm called Water-Probe that detects watermarks through well-designed prompts to the LLM. Our key motivation is that current watermarked LLMs expose consistent biases under the same watermark key, resulting in similar differences across prompts under different watermark keys. Experiments show that almost all mainstream watermarking algorithms are easily identified with our well-designed prompts, while Water-Probe demonstrates a minimal false positive rate for non-watermarked LLMs. Finally, we propose that the key to enhancing the imperceptibility of watermarked LLMs is to increase the randomness of watermark key selection. Based on this, we introduce the Water-Bag strategy, which significantly improves watermark imperceptibility by merging multiple watermark keys.

cs.CR↗

arXiv2022

Graph Neural Network and Koopman Models for Learning Networked Dynamics: A Comparative Study on Power Grid Transients Prediction

Continuous monitoring of the spatio-temporal dynamic behavior of critical infrastructure networks, such as the power systems, is a challenging but important task. In particular, accurate and timely prediction of the (electro-mechanical) transient dynamic trajectories of the power grid is necessary for early detection of any instability and prevention of catastrophic failures. Existing approaches for the prediction of dynamic trajectories either rely on the availability of accurate physical models of the system, use computationally expensive time-domain simulations, or are applicable only at local prediction problems (e.g., a single generator). In this paper, we report the application of two broad classes of data-driven learning models -- along with their algorithmic implementation and performance evaluation -- in predicting transient trajectories in power networks using only streaming measurements and the network topology as input. One class of models is based on the Koopman operator theory which allows for capturing the nonlinear dynamic behavior via an infinite-dimensional linear operator. The other class of models is based on the graph convolutional neural networks which are adept at capturing the inherent spatio-temporal correlations within the power network. Transient dynamic datasets for training and testing the models are synthesized by simulating a wide variety of load change events in the IEEE 68-bus system, categorized by the load change magnitudes, as well as by the degree of connectivity and the distance to nearest generator nodes. The results confirm that the proposed predictive models can successfully predict the post-disturbance transient evolution of the system with a high level of accuracy.

eess.SY↗

arXiv2017

A Novel Construction of Low-Complexity MDS Codes with Optimal Repair Capability for Distributed Storage Systems

Maximum-distance-separable (MDS) codes are a class of erasure codes that are widely adopted to enhance the reliability of distributed storage systems (DSS). In (n, k) MDS coded DSS, the original data are stored into n distributed nodes in an efficient manner such that each storage node only contains a small amount (i.e., 1/k) of the data and a data collector connected to any k nodes can retrieve the entire data. On the other hand, a node failure can be repaired (i.e., stored data at the failed node can be successfully recovered) by downloading data segments from other surviving nodes. In this paper, we develop a new approach to construction of simple (5, 3) MDS codes. With judiciously block-designed generator matrices, we show that the proposed MDS codes have a minimum stripe size α = 2 and can be constructed over a small (Galois) finite field F4 of only four elements, both facilitating low-complexity computations and implementations for data storage, retrieval and repair. In addition, with the proposed MDS codes, any single node failure can be repaired through interference alignment technique with a minimum data amount downloaded from the surviving nodes; i.e., the proposed codes ensure optimal exact-repair of any single node failure using the minimum bandwidth. The low-complexity and all-node-optimal-repair properties of the proposed MDS codes make them readily deployed for practical DSS.

cs.IT↗