LLM-as-Judge Framework for Evaluating Tone-Induced Hallucination in Vision-Language Models

Vision-Language Models (VLMs) are increasingly deployed in settings where reliable visual grounding carries operational consequences, yet their behavior under progressively coercive prompt phrasing remains undercharacterized. Existing hallucination benchmarks predominantly rely on neutral prompts and binary detection, leaving open how both the incidence and the intensity of fabrication respond to graded linguistic pressure across structurally distinct task types. We present Ghost-100, a procedurally constructed benchmark of 800 synthetically generated images spanning eight categories across three task families: text-illegibility, time-reading, and object-absence, each designed under a negative-ground-truth principle that guarantees the queried target is absent, illegible, or indeterminate by construction. Every image is paired with five prompts drawn from a structured 5-Level Prompt Intensity Framework, holding the image and task identity fixed while varying only directive force, so that tone is isolated as the sole independent variable. We adopt a dual-track evaluation protocol: a rule-based H-Rate measuring the proportion of responses in which a model crosses from grounded refusal into unsupported positive commitment, and a GPT-4o-mini-judged H-Score on a 1-5 scale characterizing the confidence and specificity of fabrication once it occurs. We additionally release a three-stage automated validation workflow, which retrospectively confirms 717 of 800 images as strictly compliant. Evaluating nine open-weight VLMs, we find that H-Rate and H-Score dissociate substantially across model families, reading-style and presence-detection subsets respond to prompt pressure in qualitatively different ways, and several models exhibit non-monotonic sensitivity peaking at intermediate tone levels: patterns that aggregate metrics obscure.

QuarkMedSearch: A Long-Horizon Deep Search Agent for Exploring Medical Intelligence

As agentic foundation models continue to evolve, how to further improve their performance in vertical domains has become an important challenge. To this end, building upon Tongyi DeepResearch, a powerful agentic foundation model, we focus on the Chinese medical deep search scenario and propose QuarkMedSearch, systematically exploring a full-pipeline approach spanning medical multi-hop data construction, training strategies, and evaluation benchmarks to further push and assess its performance upper bound in vertical domains. Specifically, for data synthesis, to address the scarcity of deep search training data in the medical domain, we combine a large-scale medical knowledge graph with real-time online exploration to construct long-horizon medical deep search training data; for post-training, we adopt a two-stage SFT and RL training strategy that progressively enhances the model's planning, tool invocation, and reflection capabilities required for deep search, while maintaining search efficiency; for evaluation, we collaborate with medical experts to construct the QuarkMedSearch Benchmark through rigorous manual verification. Experimental results demonstrate that QuarkMedSearch achieves state-of-the-art performance among open-source models of comparable scale on the QuarkMedSearch Benchmark, while also maintaining strong competitiveness on general benchmarks.

Toward Accountable AI-Generated Content on Social Platforms: Steganographic Attribution and Multimodal Harm Detection

The rapid growth of generative AI has introduced new challenges in content moderation and digital forensics. In particular, benign AI-generated images can be paired with harmful or misleading text, creating difficult-to-detect misuse. This contextual misuse undermines the traditional moderation framework and complicates attribution, as synthetic images typically lack persistent metadata or device signatures. We introduce a steganography enabled attribution framework that embeds cryptographically signed identifiers into images at creation time and uses multimodal harmful content detection as a trigger for attribution verification. Our system evaluates five watermarking methods across spatial, frequency, and wavelet domains. It also integrates a CLIP-based fusion model for multimodal harmful-content detection. Experiments demonstrate that spread-spectrum watermarking, especially in the wavelet domain, provides strong robustness under blur distortions, and our multimodal fusion detector achieves an AUC-ROC of 0.99, enabling reliable cross-modal attribution verification. These components form an end-to-end forensic pipeline that enables reliable tracing of harmful deployments of AI-generated imagery, supporting accountability in modern synthetic media environments. Our code is available at GitHub: https://github.com/bli1/steganography

TransGP: Task-Conditioned Transformer-Guided Genetic Programming for Multitask Dynamic Flexible Job Shop Scheduling

Hyper-heuristics have become a popular approach for solving dynamic flexible job shop scheduling (DFJSS) problems. They use gradient-free optimization techniques like Genetic Programming (GP) to evolve non-differentiable heuristics. However, conventional GP methods tend to converge slowly because they rely solely on evolutionary search to find good heuristics. Existing multitask GP methods can solve multiple tasks simultaneously and speed up the search by transferring knowledge across similar tasks. But they mostly exchange heuristic building blocks without truly generating heuristics conditioned on task information. In this paper, we aim to accelerate convergence and enable task-specific heuristic generation by incorporating a task-conditioned Transformer model. The Transformer works in two ways. First, it learns the distribution of elite heuristics, biasing the search toward promising regions of the heuristic space. Second, through conditional generation, it produces heuristics tailored to specific tasks, allowing the model to handle multiple scheduling tasks at once and improving overall optimization efficiency. Based on these ideas, we propose TransGP, a Task-Conditioned Transformer-Guided GP framework. This evolutionary paradigm integrates generative modeling with GP, enabling efficient multitask heuristic learning and knowledge transfer. We evaluate TransGP on a range of DFJSS scenarios. Experimental results show that TransGP consistently outperforms multitask GP baselines, widely used handcrafted heuristics, and the pure Transformer model, achieving faster convergence, superior solution quality, and enhanced robustness.

Tone Matters: The Impact of Linguistic Tone on Hallucination in VLMs

Vision-Language Models (VLMs) are increasingly used in safety-critical applications that require reliable visual grounding. However, these models often hallucinate details that are not present in the image to satisfy user prompts. While recent datasets and benchmarks have been introduced to evaluate systematic hallucinations in VLMs, many hallucination behaviors remain insufficiently characterized. In particular, prior work primarily focuses on object presence or absence, leaving it unclear how prompt phrasing and structural constraints can systematically induce hallucinations. In this paper, we investigate how different forms of prompt pressure influence hallucination behavior. We introduce Ghost-100, a procedurally generated dataset of synthetic scenes in which key visual details are deliberately removed, enabling controlled analysis of absence-based hallucinations. Using a structured 5-Level Prompt Intensity Framework, we vary prompts from neutral queries to toxic demands and rigid formatting constraints. We evaluate three representative open-weight VLMs: MiniCPM-V 2.6-8B, Qwen2-VL-7B, and Qwen3-VL-8B. Across all three models, hallucination rates do not increase monotonically with prompt intensity. All models exhibit reductions at higher intensity levels at different thresholds, though not all show sustained reduction under maximum coercion. These results suggest that current safety alignment is more effective at detecting semantic hostility than structural coercion, revealing model-specific limitations in handling compliance pressure. Our dataset is available at: https://github.com/bli1/tone-matters

GauSSmart: Enhanced 3D Reconstruction through 2D Foundation Models and Geometric Filtering

Scene reconstruction has emerged as a central challenge in computer vision, with approaches such as Neural Radiance Fields (NeRF) and Gaussian Splatting achieving remarkable progress. While Gaussian Splatting demonstrates strong performance on large-scale datasets, it often struggles to capture fine details or maintain realism in regions with sparse coverage, largely due to the inherent limitations of sparse 3D training data. In this work, we propose GauSSmart, a hybrid method that effectively bridges 2D foundational models and 3D Gaussian Splatting reconstruction. Our approach integrates established 2D computer vision techniques, including convex filtering and semantic feature supervision from foundational models such as DINO, to enhance Gaussian-based scene reconstruction. By leveraging 2D segmentation priors and high-dimensional feature embeddings, our method guides the densification and refinement of Gaussian splats, improving coverage in underrepresented areas and preserving intricate structural details. We validate our approach across three datasets, where GauSSmart consistently outperforms existing Gaussian Splatting in the majority of evaluated scenes. Our results demonstrate the significant potential of hybrid 2D-3D approaches, highlighting how the thoughtful combination of 2D foundational models with 3D reconstruction pipelines can overcome the limitations inherent in either approach alone.

Efficient lattice field theory simulation using adaptive normalizing flow on a resistive memory-based neural differential equation solver

Lattice field theory (LFT) simulations underpin advances in classical statistical mechanics and quantum field theory, providing a unified computational framework across particle, nuclear, and condensed matter physics. However, the application of these methods to high-dimensional systems remains severely constrained by several challenges, including the prohibitive computational cost and limited parallelizability of conventional sampling algorithms such as hybrid Monte Carlo (HMC), the substantial training expense associated with traditional normalizing flow models, and the inherent energy inefficiency of digital hardware architectures. Here, we introduce a software-hardware co-design that integrates an adaptive normalizing flow (ANF) model with a resistive memory-based neural differential equation solver, enabling efficient generation of LFT configurations. Software-wise, ANF enables efficient parallel generation of statistically independent configurations, thereby reducing computational costs, while low-rank adaptation (LoRA) allows cost-effective fine-tuning across diverse simulation parameters. Hardware-wise, in-memory computing with resistive memory substantially enhances both parallelism and energy efficiency. We validate our approach on the scalar phi4 theory and the effective field theory of graphene wires, using a hybrid analog-digital neural differential equation solver equipped with a 180 nm resistive memory in-memory computing macro. Our co-design enables low-cost computation, achieving approximately 8.2-fold and 13.9-fold reductions in integrated autocorrelation time over HMC, while requiring fine-tuning of less than 8% of the weights via LoRA. Compared to state-of-the-art GPUs, our co-design achieves up to approximately 16.1- and 17.0-fold speedups for the two tasks, as well as 73.7- and 138.0-fold improvements in energy efficiency.

Quality Diversity Genetic Programming for Learning Scheduling Heuristics

Real-world optimization often demands diverse, high-quality solutions. Quality-Diversity (QD) optimization is a multifaceted approach in evolutionary algorithms that aims to generate a set of solutions that are both high-performing and diverse. QD algorithms have been successfully applied across various domains, providing robust solutions by exploring diverse behavioral niches. However, their application has primarily focused on static problems, with limited exploration in the context of dynamic combinatorial optimization problems. Furthermore, the theoretical understanding of QD algorithms remains underdeveloped, particularly when applied to learning heuristics instead of directly learning solutions in complex and dynamic combinatorial optimization domains, which introduces additional challenges. This paper introduces a novel QD framework for dynamic scheduling problems. We propose a map-building strategy that visualizes the solution space by linking heuristic genotypes to their behaviors, enabling their representation on a QD map. This map facilitates the discovery and maintenance of diverse scheduling heuristics. Additionally, we conduct experiments on both fixed and dynamically changing training instances to demonstrate how the map evolves and how the distribution of solutions unfolds over time. We also discuss potential future research directions that could enhance the learning process and broaden the applicability of QD algorithms to dynamic combinatorial optimization challenges.

Detecting Manipulated Contents Using Knowledge-Grounded Inference

The detection of manipulated content, a prevalent form of fake news, has been widely studied in recent years. While existing solutions have been proven effective in fact-checking and analyzing fake news based on historical events, the reliance on either intrinsic knowledge obtained during training or manually curated context hinders them from tackling zero-day manipulated content, which can only be recognized with real-time contextual information. In this work, we propose Manicod, a tool designed for detecting zero-day manipulated content. Manicod first sources contextual information about the input claim from mainstream search engines, and subsequently vectorizes the context for the large language model (LLM) through retrieval-augmented generation (RAG). The LLM-based inference can produce a "truthful" or "manipulated" decision and offer a textual explanation for the decision. To validate the effectiveness of Manicod, we also propose a dataset comprising 4270 pieces of manipulated fake news derived from 2500 recent real-world news headlines. Manicod achieves an overall F1 score of 0.856 on this dataset and outperforms existing methods by up to 1.9x in F1 score on their benchmarks on fact-checking and claim verification.

KSHSeek: Data-Driven Approaches to Mitigating and Detecting Knowledge-Shortcut Hallucinations in Generative Models

The emergence of large language models (LLMs) has significantly advanced the development of natural language processing (NLP), especially in text generation tasks like question answering. However, model hallucinations remain a major challenge in natural language generation (NLG) tasks due to their complex causes. We systematically expand on the causes of factual hallucinations from the perspective of knowledge shortcuts, analyzing hallucinations arising from correct and defect-free data and demonstrating that knowledge-shortcut hallucinations are prevalent in generative models. To mitigate this issue, we propose a high similarity pruning algorithm at the data preprocessing level to reduce spurious correlations in the data. Additionally, we design a specific detection method for knowledge-shortcut hallucinations to evaluate the effectiveness of our mitigation strategy. Experimental results show that our approach effectively reduces knowledge-shortcut hallucinations, particularly in fine-tuning tasks, without negatively impacting model performance in question answering. This work introduces a new paradigm for mitigating specific hallucination issues in generative models, enhancing their robustness and reliability in real-world applications.