arXiv subjects

Binghao Wang

Publications and source records attributed to Binghao Wang.

5 recordsLinked to original sources

arXiv2026

Revis: Sparse Latent Steering to Mitigate Object Hallucination in Large Vision-Language Models

Despite the advanced capabilities of Large Vision-Language Models (LVLMs), they frequently suffer from object hallucination. One reason is that visual features and pretrained textual representations often become intertwined in the deeper network layers. To address this, we propose REVIS, a training-free framework designed to explicitly re-activate this suppressed visual information. Rooted in latent space geometry, REVIS extracts the pure visual information vector via orthogonal projection and employs a calibrated strategy to perform sparse intervention only at the precise depth where suppression occurs. This surgical approach effectively restores visual information with minimal computational cost. Empirical evaluations on standard benchmarks demonstrate that REVIS reduces object hallucination rates by approximately 19% compared to state-of-the-art baselines, while preserving general reasoning capabilities.

cs.AI↗

arXiv2026

Light Alignment Improves LLM Safety via Model Self-Reflection with a Single Neuron

The safety of large language models (LLMs) has increasingly emerged as a fundamental aspect of their development. Existing safety alignment for LLMs is predominantly achieved through post-training methods, which are computationally expensive and often fail to generalize well across different models. A small number of lightweight alignment approaches either rely heavily on prior-computed safety injections or depend excessively on the model's own capabilities, resulting in limited generalization and degraded efficiency and usability during generation. In this work, we propose a safety-aware decoding method that requires only low-cost training of an expert model and employs a single neuron as a gating mechanism. By effectively balancing the model's intrinsic capabilities with external guidance, our approach simultaneously preserves utility and enhances output safety. It demonstrates clear advantages in training overhead and generalization across model scales, offering a new perspective on lightweight alignment for the safe and practical deployment of large language models. Code: https://github.com/Beijing-AISI/NGSD.

cs.AI↗

arXiv2025

MedKGEval: A Knowledge Graph-Based Multi-Turn Evaluation Framework for Open-Ended Patient Interactions with Clinical LLMs

The reliable evaluation of large language models (LLMs) in medical applications remains an open challenge, particularly in capturing the complexity of multi-turn doctor-patient interactions that unfold in real clinical environments. Existing evaluation methods typically rely on post hoc review of full conversation transcripts, thereby neglecting the dynamic, context-sensitive nature of medical dialogues and the evolving informational needs of patients. In this work, we present MedKGEval, a novel multi-turn evaluation framework for clinical LLMs grounded in structured medical knowledge. Our approach introduces three key contributions: (1) a knowledge graph-driven patient simulation mechanism, where a dedicated control module retrieves relevant medical facts from a curated knowledge graph, thereby endowing the patient agent with human-like and realistic conversational behavior. This knowledge graph is constructed by integrating open-source resources with additional triples extracted from expert-annotated datasets; (2) an in-situ, turn-level evaluation framework, where each model response is assessed by a Judge Agent for clinical appropriateness, factual correctness, and safety as the dialogue progresses using a suite of fine-grained, task-specific metrics; (3) a comprehensive multi-turn benchmark of eight state-of-the-art LLMs, demonstrating MedKGEval's ability to identify subtle behavioral flaws and safety risks that are often overlooked by conventional evaluation pipelines. Although initially designed for Chinese and English medical applications, our framework can be readily extended to additional languages by switching the input knowledge graphs, ensuring seamless bilingual support and domain-specific applicability.

cs.AI↗

arXiv2025

GenPTW: Latent Image Watermarking for Provenance Tracing and Tamper Localization

The proliferation of generative image models has revolutionized AIGC creation while amplifying concerns over content provenance and manipulation forensics. Existing methods are typically either unable to localize tampering or restricted to specific generative settings, limiting their practical utility. We propose \textbf{GenPTW}, a \textbf{Gen}eral watermarking framework that unifies \textbf{P}rovenance tracing and \textbf{T}amper localization in latent space. It supports both in-generation and post-generation embedding without altering the generative process, and is plug-and-play compatible with latent diffusion models (LDMs) and visual autoregressive (VAR) models. To achieve precise provenance tracing and tamper localization, we embed the watermark using two complementary mechanisms: cross-attention fusion aligned with latent semantics and spatial fusion providing explicit spatial guidance for edit sensitivity. A tamper-aware extractor jointly conducts provenance tracing and tamper localization by leveraging watermark features together with high-frequency features. Experiments show that GenPTW maintains high visual fidelity and strong robustness against diverse AIGC-editing.

cs.CR↗

arXiv2021

Efficient Trajectory Compression and Range Query Processing

Nowadays, there are ubiquitousness of GPS sensors in various devices collecting, transmitting and storing tremendous trajectory data. However, such an unprecedented scale of GPS data has posed an urgent demand for not only an effective storage mechanism but also an efficient query mechanism. Line simplification in online mode, searving as a mainstream trajectory compression method, plays an important role to attack this issue. But for the existing algorithms, either their time cost is extremely high, or the accuracy loss after the compression is completely unacceptable. To attack this issue, we propose $ε\_$Region based Online trajectory Compression with Error bounded (ROCE for short), which makes the best balance among the accuracy loss, the time cost and the compression rate. The range query serves as a primitive, yet quite essential operation on analyzing trajectories. Each trajectory is usually seen as a sequence of discrete points, and in most previous work, a trajectory is judged to be overlapped with the query region R iff there is at least one point in this trajectory falling in R. But this traditional criteria is not suitable when the queried trajectories are compressed, because there may be hundreds of points discarded between each two adjacent points and the points in each compressed trajectory are quite sparse. And many trajectories could be missing in the result set. To address this, in this paper, a new criteria based on the probability and an efficient Range Query processing algorithm on Compressed trajectories RQC are proposed. In addition, an efficient index \emph{ASP\_tree} and lots of novel techniques are also presented to accelerate the processing of trajectory compression and range queries obviously. Extensive experiments have been done on multiple real datasets, and the results demonstrate superior performance of our methods.

cs.DB↗