Moreover, 33.7% of the questions are cross-page questions

Specifically, as part of the question, it contains a referring query that references related video contexts, called referred context .

Multimodal Large Language Models (MLLMs) have shown significant potential in surgical video understanding. With improved zero-shot performance and more effective human-machine interaction, they provide a strong foundation for advancing surgical education and assistance. However, existing research and datasets primarily focus on understanding surgical procedures and workflows, while paying limited attention to the critical role of anatomical comprehension. In clinical practice, surgeons rely heavily on precise anatomical understanding to interpret, review, and learn from surgical videos. To fill this gap, we introduce the Neurosurgical Anatomy Benchmark (NeuroABench), the first multimodal benchmark explicitly created to evaluate anatomical understanding in the neurosurgical domain. NeuroABench consists of 9 hours of annotated neurosurgical videos covering 89 distinct procedures and is developed using a novel multimodal annotation pipeline with multiple review cycles. The benchmark evaluates the identification of 68 clinical anatomical structures, providing a rigorous and standardized framework for assessing model performance. Experiments on over 10 state-of-the-art MLLMs reveal significant limitations, with the best-performing model achieving only 40.87% accuracy in anatomical identification tasks. To further evaluate the benchmark, we extract a subset of the dataset and conduct an informative test with four neurosurgical trainees. The results show that the best-performing student achieves 56% accuracy, with the lowest scores of 28% and an average score of 46.5%. While the best MLLM performs comparably to the lowest-scoring student, it still lags significantly behind the group's average performance. This comparison underscores both the progress of MLLMs in anatomical understanding and the substantial gap that remains in achieving human-level performance.

Abstract--Few-shot prompting has emerged as a practical alternative to fine-tuning for leveraging the capabilities of large language models (LLMs) in specialized tasks. However, its effectiveness depends heavily on the selection and quality of in-context examples, particularly in complex domains. In this work, we examine retrieval-augmented prompting as a strategy to improve few-shot performance in code vulnerability detection, where the goal is to identify one or more security-relevant weaknesses present in a given code snippet from a predefined set of vulnerability categories. We perform a systematic evaluation using the Gemini-1.5-Flash Our results show that retrieval-augmented prompting consistently outperforms the other prompting strategies. At 20 shots, it achieves an F1 score of 74.05% and a partial match accuracy of 83.90%. We further compare this approach against zero-shot prompting and several fine-tuned models, including Gemini-1.5-Flash Retrieval-augmented prompting outperforms both zero-shot (F1 score: 36.35%, On the other hand, fine-tuning CodeBERT yields higher performance (F1 score: 91.22%, partial match accuracy: 91.30%) but requires additional training, maintenance effort, and resources.

Safety evaluation of multimodal foundation models often treats vision and language inputs separately, missing risks from joint interpretation where benign content becomes harmful in combination. Existing approaches also fail to distinguish clearly unsafe content from borderline cases, leading to problematic over-blocking or under-refusal of genuinely harmful content. We present Vision Language Safety Understanding (VLSU), a comprehensive framework to systematically evaluate multimodal safety through fine-grained severity classification and combinatorial analysis across 17 distinct safety patterns. Using a multi-stage pipeline with real-world images and human annotation, we construct a large-scale benchmark of 8,187 samples spanning 15 harm categories. Our evaluation of eleven state-of-the-art models reveals systematic joint understanding failures: while models achieve 90%-plus accuracy on clear unimodal safety signals, performance degrades substantially to 20-55% when joint image-text reasoning is required to determine the safety label. Most critically, 34% of errors in joint image-text safety classification occur despite correct classification of the individual modalities, further demonstrating absent compositional reasoning capabilities. Additionally, we find that models struggle to balance refusing unsafe content while still responding to borderline cases that deserve engagement. For example, we find that instruction framing can reduce the over-blocking rate on borderline content from 62.4% to 10.4% in Gemini-1.5, but only at the cost of under-refusing on unsafe content with refusal rate dropping from 90.8% to 53.9%. Overall, our framework exposes weaknesses in joint image-text understanding and alignment gaps in current models, and provides a critical test bed to enable the next milestones in research on robust vision-language safety.

We introduce WearVQA, the first benchmark specifically designed to evaluate the Visual Question Answering (VQA) capabilities of multi-modal AI assistant on wearable devices like smart glasses. Unlike prior benchmarks that focus on high-quality, third-person imagery, WearVQA reflects the unique challenges of egocentric interaction--where visual inputs may be occluded, poorly lit, unzoomed, or blurry, and questions are grounded in realistic wearable use cases. The benchmark comprises 2,520 carefully curated image-question-answer triplets, spanning 7 diverse image domains including both text-centric and general scenes, 10 cognitive task types ranging from basic recognition to various forms of reasoning, and 6 common wearables-specific image quality issues. All questions are designed to be answerable using only the visual input and common senses. WearVQA is paired with a rigorous LLM-as-a-judge evaluation framework with 96% labeling accuracy. Open-source and proprietary multi-modal LLMs achieved a QA accuracy as low as 24-52% on WearVQA, with substantial drops on lower-quality images and reasoning-heavy tasks.

We introduce Speech-based Intelligence Quotient (SIQ) as a new form of human cognition-inspired evaluation pipeline for voice understanding large language models, LLM Voice, designed to assess their voice understanding ability. Moving beyond popular voice understanding metrics such as word error rate (WER), SIQ examines LLM Voice across three cognitive levels motivated by Bloom's Taxonomy: (1) Remembering (i.e., WER for verbatim accuracy); (2) Understanding (i.e., similarity of LLM's interpretations); and (3) Application (i.e., QA accuracy for simulating downstream tasks). We demonstrate that SIQ not only quantifies voice understanding abilities but also provides unified comparisons between cascaded methods (e.g., ASR LLM) and end-to-end models, identifies annotation errors in existing benchmarks, and detects hallucinations in LLM Voice. Our framework represents a first-of-its-kind intelligence examination that bridges cognitive principles with voice-oriented benchmarks, while exposing overlooked challenges in multi-modal training. Our code and data will be open source to encourage future studies.

Automatic post-editing (APE) aims to correct errors in machine-translated text, enhancing translation quality, while reducing the need for human intervention. Despite advances in neural machine translation (NMT), the development of effective APE systems has been hindered by the lack of large-scale multilingual datasets specifically tailored to NMT outputs. To address this gap, we present and release LangMark, a new human-annotated multilingual APE dataset for English translation to seven languages: Brazilian Portuguese, French, German, Italian, Japanese, Russian, and Spanish. The dataset has 206,983 triplets, with each triplet consisting of a source segment, its NMT output, and a human post-edited translation. Annotated by expert human linguists, our dataset offers both linguistic diversity and scale. Leveraging this dataset, we empirically show that Large Language Models (LLMs) with few-shot prompting can effectively perform APE, improving upon leading commercial and even proprietary machine translation systems. We believe that this new resource will facilitate the future development and evaluation of APE systems.