Quantized Vision-Language Models for Damage Assessment: A Comparative Study of LLaVA-1.5-7B Quantization Levels

Bridge infrastructure inspection is a critical but labor-intensive task requiring expert assessment of structural damage such as rebar exposure, cracking, and corrosion. This paper presents a comprehensive study of quantized Vision-Language Models (VLMs) for automated bridge damage assessment, focusing on the trade-offs between description quality, inference speed, and resource requirements. We develop an end-to-end pipeline combining LLaVA-1.5-7B for visual damage analysis, structured JSON extraction, and rule-based priority scoring. To enable deployment on consumer-grade GPUs, we conduct a systematic comparison of three quantization levels: Q4_K_M, Q5_K_M, and Q8\_0 across 254 rebar exposure images. We introduce a 5-point quality evaluation framework assessing damage type recognition, severity classification. Our results demonstrate that Q5_K_M achieves the optimal balance: quality score 3.18$\pm$1.35/5.0, inference time 5.67s/image, and 0.56 quality/sec efficiency -- 8.5% higher quality than Q4_K_M with only 4.5% speed reduction, while matching Q8_0's quality with 25% faster inference. Statistical analysis reveals Q5_K_M exhibits the weakest text-quality correlation (-0.148), indicating consistent performance regardless of description length.

SozKZ: Training Efficient Small Language Models for Kazakh from Scratch

Kazakh, a Turkic language spoken by over 22 million people, remains underserved by existing multilingual language models, which allocate minimal capacity to low-resource languages and employ tokenizers ill-suited to agglutinative morphology. We present SozKZ, a family of Llama-architecture language models (50M-600M parameters) trained entirely from scratch on 9 billion tokens of Kazakh text with a dedicated 50K BPE tokenizer. We evaluate all models on three Kazakh benchmarks -- multiple-choice cultural QA, reading comprehension (Belebele), and topic classification (SIB-200) -- alongside five multilingual baselines ranging from 500M to 3B parameters. Our 600M model achieves 30.3% accuracy on Kazakh cultural QA, approaching the 32.0% of Llama-3.2-1B (2x larger), and 25.5% on SIB-200 topic classification, surpassing all evaluated multilingual models up to 2B parameters. We observe consistent scaling from 50M to 600M, with MC QA accuracy rising from 22.8% to 30.3%, suggesting that further scaling remains beneficial. These results demonstrate that small, dedicated models trained from scratch with a language-appropriate tokenizer offer a viable path for low-resource language technology, achieving competitive performance at a fraction of the computational cost. All models and the tokenizer are released under open licenses.

LogicDiff: Logic-Guided Denoising Improves Reasoning in Masked Diffusion Language Models

Masked diffusion language models (MDLMs) generate text by iteratively unmasking tokens from a fully masked sequence, offering parallel generation and bidirectional context. However, their standard confidence-based unmasking strategy systematically defers high-entropy logical connective tokens, the critical branching points in reasoning chains, leading to severely degraded reasoning performance. We introduce LogicDiff, an inference-time method that replaces confidence-based unmasking with logic-role-guided unmasking. A lightweight classification head (4.2M parameters, 0.05% of the base model) predicts the logical role of each masked position (premise, connective, derived step, conclusion, or filler) from the base model's hidden states with 98.4% accuracy. A dependency-ordered scheduler then unmasks tokens in logical dependency order: premises first, then connectives, then derived steps, then conclusions. Without modifying a single parameter of the base model and without any reinforcement learning or task-specific training, LogicDiff improves LLaDA-8B-Instruct accuracy from 22.0% to 60.7% on GSM8K (+38.7 percentage points) and from 23.6% to 29.2% on MATH-500 (+5.6 pp), with less than 6% speed overhead. Our results demonstrate that a substantial portion of the reasoning deficit in MDLMs is attributable to suboptimal token unmasking order, not to limitations of the model's learned representations.

WiFi2Cap: Semantic Action Captioning from Wi-Fi CSI via Limb-Level Semantic Alignment

Privacy-preserving semantic understanding of human activities is important for indoor sensing, yet existing Wi-Fi CSI-based systems mainly focus on pose estimation or predefined action classification rather than fine-grained language generation. Mapping CSI to natural-language descriptions remains challenging because of the semantic gap between wireless signals and language and direction-sensitive ambiguities such as left/right limb confusion. We propose WiFi2Cap, a three-stage framework for generating action captions directly from Wi-Fi CSI. A vision-language teacher learns transferable supervision from synchronized video-text pairs, and a CSI student is aligned to the teacher's visual space and text embeddings. To improve direction-sensitive captioning, we introduce a Mirror-Consistency Loss that reduces mirrored-action and left-right ambiguities during cross-modal alignment. A prefix-tuned language model then generates action descriptions from CSI embeddings. We also introduce the WiFi2Cap Dataset, a synchronized CSI-RGB-sentence benchmark for semantic captioning from Wi-Fi signals. Experimental results show that WiFi2Cap consistently outperforms baseline methods on BLEU-4, METEOR, ROUGE-L, CIDEr, and SPICE, demonstrating effective privacy-friendly semantic sensing.

Tabular LLMs for Interpretable Few-Shot Alzheimer's Disease Prediction with Multimodal Biomedical Data

Accurate diagnosis of Alzheimer's disease (AD) requires handling tabular biomarker data, yet such data are often small and incomplete, where deep learning models frequently fail to outperform classical methods. Pretrained large language models (LLMs) offer few-shot generalization, structured reasoning, and interpretable outputs, providing a powerful paradigm shift for clinical prediction. We propose TAP-GPT Tabular Alzheimer's Prediction GPT, a domain-adapted tabular LLM framework built on TableGPT2 and fine-tuned for few-shot AD classification using tabular prompts rather than plain texts. We evaluate TAP-GPT across four ADNI-derived datasets, including QT-PAD biomarkers and region-level structural MRI, amyloid PET, and tau PET for binary AD classification. Across multimodal and unimodal settings, TAP-GPT improves upon its backbone models and outperforms traditional machine learning baselines in the few-shot setting while remaining competitive with state-of-the-art general-purpose LLMs. We show that feature selection mitigates degradation in high-dimensional inputs and that TAP-GPT maintains stable performance under simulated and real-world missingness without imputation. Additionally, TAP-GPT produces structured, modality-aware reasoning aligned with established AD biology and shows greater stability under self-reflection, supporting its use in iterative multi-agent systems. To our knowledge, this is the first systematic application of a tabular-specialized LLM to multimodal biomarker-based AD prediction, demonstrating that such pretrained models can effectively address structured clinical prediction tasks and laying the foundation for tabular LLM-driven multi-agent clinical decision-support systems. The source code is publicly available on GitHub: https://github.com/sophie-kearney/TAP-GPT.

Consistent but Dangerous: Per-Sample Safety Classification Reveals False Reliability in Medical Vision-Language Models

Consistency under paraphrase, the property that semantically equivalent prompts yield identical predictions, is increasingly used as a proxy for reliability when deploying medical vision-language models (VLMs). We show this proxy is fundamentally flawed: a model can achieve perfect consistency by relying on text patterns rather than the input image. We introduce a four-quadrant per-sample safety taxonomy that jointly evaluates consistency (stable predictions across paraphrased prompts) and image reliance (predictions that change when the image is removed). Samples are classified as Ideal (consistent and image-reliant), Fragile (inconsistent but image-reliant), Dangerous (consistent but not image-reliant), or Worst (inconsistent and not image-reliant). Evaluating five medical VLM configurations across two chest X-ray datasets (MIMIC-CXR, PadChest), we find that LoRA fine-tuning dramatically reduces flip rates but shifts a majority of samples into the Dangerous quadrant: LLaVA-Rad Base achieves a 1.5% flip rate on PadChest while 98.5% of its samples are Dangerous. Critically, Dangerous samples exhibit high accuracy (up to 99.6%) and low entropy, making them invisible to standard confidence-based screening. We observe a negative correlation between flip rate and Dangerous fraction (r = -0.89, n=10) and recommend that deployment evaluations always pair consistency checks with a text-only baseline: a single additional forward pass that exposes the false reliability trap.

MzansiText and MzansiLM: An Open Corpus and Decoder-Only Language Model for South African Languages

Decoder-only language models can be adapted to diverse tasks through instruction finetuning, but the extent to which this generalizes at small scale for low-resource languages remains unclear. We focus on the languages of South Africa, where we are not aware of a publicly available decoder-only model that explicitly targets all eleven official written languages, nine of which are low-resource. We introduce MzansiText, a curated multilingual pretraining corpus with a reproducible filtering pipeline, and MzansiLM, a 125M-parameter language model trained from scratch. We evaluate MzansiLM on natural language understanding and generation using three adaptation regimes: monolingual task-specific finetuning, multilingual task-specific finetuning, and general multi-task instruction finetuning. Monolingual task-specific finetuning achieves strong performance on data-to-text generation, reaching 20.65 BLEU on isiXhosa and competing with encoder-decoder baselines over ten times larger. Multilingual task-specific finetuning benefits closely related languages on topic classification, achieving 78.5% macro-F1 on isiXhosa news classification. While MzansiLM adapts effectively to supervised NLU and NLG tasks, few-shot reasoning remains challenging at this model size, with performance near chance even for much larger decoder-only models. We release MzansiText and MzansiLM to provide a reproducible decoder-only baseline and clear guidance on adaptation strategies for South African languages at small scale.

More Than Sum of Its Parts: Deciphering Intent Shifts in Multimodal Hate Speech Detection

Combating hate speech on social media is critical for securing cyberspace, yet relies heavily on the efficacy of automated detection systems. As content formats evolve, hate speech is transitioning from solely plain text to complex multimodal expressions, making implicit attacks harder to spot. Current systems, however, often falter on these subtle cases, as they struggle with multimodal content where the emergent meaning transcends the aggregation of individual modalities. To bridge this gap, we move beyond binary classification to characterize semantic intent shifts where modalities interact to construct implicit hate from benign cues or neutralize toxicity through semantic inversion. Guided by this fine-grained formulation, we curate the Hate via Vision-Language Interplay (H-VLI) benchmark where the true intent hinges on the intricate interplay of modalities rather than overt visual or textual slurs. To effectively decipher these complex cues, we further propose the Asymmetric Reasoning via Courtroom Agent DEbate (ARCADE) framework. By simulating a judicial process where agents actively argue for accusation and defense, ARCADE forces the model to scrutinize deep semantic cues before reaching a verdict. Extensive experiments demonstrate that ARCADE significantly outperforms state-of-the-art baselines on H-VLI, particularly for challenging implicit cases, while maintaining competitive performance on established benchmarks. Our code and data are available at: https://github.com/Sayur1n/H-VLI

Parameter-efficient Prompt Tuning and Hierarchical Textual Guidance for Few-shot Whole Slide Image Classification

Whole Slide Images (WSIs) are giga-pixel in scale and are typically partitioned into small instances in WSI classification pipelines for computational feasibility. However, obtaining extensive instance level annotations is costly, making few-shot weakly supervised WSI classification (FSWC) crucial for learning from limited slide-level labels. Recently, pre-trained vision-language models (VLMs) have been adopted in FSWC, yet they exhibit several limitations. Existing prompt tuning methods in FSWC substantially increase both the number of trainable parameters and inference overhead. Moreover, current methods discard instances with low alignment to text embeddings from VLMs, potentially leading to information loss. To address these challenges, we propose two key contributions. First, we introduce a new parameter efficient prompt tuning method by scaling and shifting features in text encoder, which significantly reduces the computational cost. Second, to leverage not only the pre-trained knowledge of VLMs, but also the inherent hierarchical structure of WSIs, we introduce a WSI representation learning approach with a soft hierarchical textual guidance strategy without utilizing hard instance filtering. Comprehensive evaluations on pathology datasets covering breast, lung, and ovarian cancer types demonstrate consistent improvements up-to 10.9%, 7.8%, and 13.8% respectively, over the state-of-the-art methods in FSWC. Our method reduces the number of trainable parameters by 18.1% on both breast and lung cancer datasets, and 5.8% on the ovarian cancer dataset, while also excelling at weakly-supervised tumor localization. Code at https://github.com/Jayanie/HIPSS.

Edit Spillover as a Probe: Do Image Editing Models Implicitly Understand World Relations?

Instruction-following image editing models are expected to modify only the specified region while keeping the rest of the image unchanged. However, in practice, we observe a pervasive phenomenon -- edit spillover: models alter semantically related but unspecified content outside the edit region. This raises a fundamental question -- does spillover reflect genuine implicit world understanding, or is it merely attention leakage? We propose EditSpilloverProbe, a systematic framework that repurposes edit spillover as a natural probe for world knowledge in image editing models. We introduce a spillover taxonomy (spatial, semantic, mixed, random), an automated detection-and-classification pipeline, and a benchmark dataset constructed from real-world Chinese text editing tasks, EditSpilloverBench. Systematic evaluation of 5 representative editing models reveals three core findings: (1) spillover rates vary dramatically across architectures, from 3.49% to 11.46%, with a 3.3x ratio; (2) absolute semantic spillover quantity reveals models' world understanding capability -- nano_banana produces the most semantic spillover (27.8 per image), while qwen_2511 has the most precise editing control but lower semantic spillover (16.3 per image), revealing a trade-off between editing control and world understanding; (3) spatial decay analysis shows spillover area density decays exponentially with distance, but the proportion of semantically relevant spillover remains constant (40%-58%), providing direct evidence that semantic spillover reflects genuine world understanding rather than spatial diffusion.