Virtual screening plays a critical role in modern drug discovery by enabling the identification of promising candidate molecules for experimental validation. Traditional machine learning methods such, as Support Vector Machines (SVM) and XGBoost, rely on predefined molecular representations, often leading to information loss and potential bias. In contrast, deep learning approaches-particularly Graph Convolutional Networks (GCNs)-offer a more expressive and unbiased alternative by operating directly on molecular graphs. Meanwhile, Large Language Models (LLMs) have recently demonstrated state-of-the-art performance in drug design, thanks to their capacity to capture complex chemical patterns from large-scale data via attention mechanisms. In this paper, we propose a hybrid architecture that integrates GCNs with LLM-derived embeddings to combine localized structural learning with global chemical knowledge. The LLM embeddings can be precomputed and stored in a molecular feature library, removing the need to rerun the LLM during training or inference and thus maintaining computational efficiency. We found that concatenating the LLM embeddings after each GCN layer-rather than only at the final layer-significantly improves performance, enabling deeper integration of global context throughout the network. The resulting model achieves superior results, with an F1-score of (88.8\%), outperforming standalone GCN (87.9%), XGBoost (85.5%), and SVM (85.4%) baselines.

Even in the era of rapid advances in large models, video understanding remains a highly challenging task. Compared to texts or images, videos commonly contain more information with redundancy, requiring large models to properly allocate attention at a global level for comprehensive and accurate understanding. To address this, we propose a Multimodal hierarchical Attention focusing Self-reflective Reasoning (MASR) framework for agent-based video understanding. The key innovation lies in its ability to detect and prioritize segments of videos that are highly relevant to the query. Firstly, MASR realizes Multimodal Coarse-to-fine Relevance Sensing (MCRS) which enhances the correlation between the acquired contextual information and the query. Secondly, MASR employs Dilated Temporal Expansion (DTE) to mitigate the risk of missing crucial details when extracting semantic information from the focused frames selected through MCRS. By iteratively applying MCRS and DTE in the self-reflective reasoning process, MASR is able to adaptively adjust the attention to extract highly query-relevant context and therefore improve the response accuracy. In the EgoSchema dataset, MASR achieves a remarkable 5% performance gain over previous leading approaches. In the Next-QA and IntentQA datasets, it outperforms the state-of-the-art standards by 0.2% and 0.3% respectively. In the Video-MME dataset that contains long-term videos, MASR also performs better than other agent-based methods.

Vision science has always developed models at smaller scales than the frontier of artificial intelligence. This is partially because of the academic roots of vision science, partially because of a well-founded desire to lean on reductionism to truly understand how vision works, and partially because attempts at incorporating biological inspiration into DNNs have been hamstrung by implementations that are poorly suited for GPUs. For example, most attempts at biologically-inspired DNNs have focused on inducing architectural constraints like recurrence [34,61,116-118] and different forms of feedback [59,60] that are not explicitly included in DNNs but known to play key roles in primate vision [119-123] . While we believe these approaches are important for Neuroscience and especially for constraining model hypothesis spaces in small data settings, the methods used for implementation are undeniably challenging to scale [118,124] and it is possible that the induced computational strategies could be learned by a less-constrained DNN trained with the "right" data and objective [125] . Thus, it may be that a more effective approach to reverse-engineering vision than hand-designing small-scale recurrent DNNs could be to train DNNs at large scales with approximations of the kinds of data and routines that shape biological visual systems.

Quality and diversity are two critical metrics for the training data of large language models (LLMs), positively impacting performance. Existing studies often optimize these metrics separately, typically by first applying quality filtering and then adjusting data proportions. However, these approaches overlook the inherent trade-off between quality and diversity, necessitating their joint consideration. Given a fixed training quota, it is essential to evaluate both the quality of each data point and its complementary effect on the overall dataset. In this paper, we introduce a unified data selection framework called QuaDMix, which automatically optimizes the data distribution for LLM pretraining while balancing both quality and diversity. Specifically, we first propose multiple criteria to measure data quality and employ domain classification to distinguish data points, thereby measuring overall diversity. QuaDMix then employs a unified parameterized data sampling function that determines the sampling probability of each data point based on these quality and diversity related labels. To accelerate the search for the optimal parameters involved in the QuaDMix framework, we conduct simulated experiments on smaller models and use LightGBM for parameters searching, inspired by the RegMix method. Our experiments across diverse models and datasets demonstrate that QuaDMix achieves an average performance improvement of 7.2% across multiple benchmarks. These results outperform the independent strategies for quality and diversity, highlighting the necessity and ability to balance data quality and diversity.

The rapid advancement of large language models (LLMs) has reshaped the landscape of machine translation, yet challenges persist in preserving poetic intent, cultural heritage, and handling specialized terminology in Chinese-English translation. This study constructs a diverse corpus encompassing Chinese scientific terminology, historical translation paradoxes, and literary metaphors. Utilizing a back-translation and Friedman test-based evaluation system (BT-Fried), we evaluate BLEU, CHRF, TER, and semantic similarity metrics across six major LLMs (e.g., GPT-4.5, DeepSeek V3) and three traditional translation tools. Key findings include: (1) Scientific abstracts often benefit from back-translation, while traditional tools outperform LLMs in linguistically distinct texts; (2) LLMs struggle with cultural and literary retention, exemplifying the "paradox of poetic intent"; (3) Some models exhibit "verbatim back-translation", reflecting emergent memory behavior; (4) A novel BLEU variant using Jieba segmentation and n-gram weighting is proposed. The study contributes to the empirical evaluation of Chinese NLP performance and advances understanding of cultural fidelity in AI-mediated translation.

The rapid development of generative artificial intelligence has made text to video models essential for building future multimodal world simulators. However, these models remain vulnerable to jailbreak attacks, where specially crafted prompts bypass safety mechanisms and lead to the generation of harmful or unsafe content. Such vulnerabilities undermine the reliability and security of simulation based applications. In this paper, we propose T2VShield, a comprehensive and model agnostic defense framework designed to protect text to video models from jailbreak threats. Our method systematically analyzes the input, model, and output stages to identify the limitations of existing defenses, including semantic ambiguities in prompts, difficulties in detecting malicious content in dynamic video outputs, and inflexible model centric mitigation strategies. T2VShield introduces a prompt rewriting mechanism based on reasoning and multimodal retrieval to sanitize malicious inputs, along with a multi scope detection module that captures local and global inconsistencies across time and modalities. The framework does not require access to internal model parameters and works with both open and closed source systems. Extensive experiments on five platforms show that T2VShield can reduce jailbreak success rates by up to 35 percent compared to strong baselines. We further develop a human centered audiovisual evaluation protocol to assess perceptual safety, emphasizing the importance of visual level defense in enhancing the trustworthiness of next generation multimodal simulators.

Multi-view understanding, the ability to reconcile visual information across diverse viewpoints for effective navigation, manipulation, and 3D scene comprehension, is a fundamental challenge in Multi-Modal Large Language Models (MLLMs) to be used as embodied agents. While recent MLLMs have shown impressive advances in high-level reasoning and planning, they frequently fall short when confronted with multi-view geometric consistency and cross-view correspondence. To comprehensively evaluate the challenges of MLLMs in multi-view scene reasoning, we propose All-Angles Bench, a benchmark of over 2,100 human carefully annotated multi-view question-answer pairs across 90 diverse real-world scenes. Our six tasks (counting, attribute identification, relative distance, relative direction, object manipulation, and camera pose estimation) specifically test model's geometric correspondence and the capacity to align information consistently across views. Our extensive experiments, benchmark on 27 representative MLLMs including Gemini-2.0-Flash, Claude-3.7-Sonnet, and GPT-4o against human evaluators reveals a substantial performance gap, indicating that current MLLMs remain far from human-level proficiency. Through in-depth analysis, we show that MLLMs are particularly underperforming under two aspects: (1) cross-view correspondence for partially occluded views and (2) establishing the coarse camera poses. These findings highlight the necessity of domain-specific refinements or modules that embed stronger multi-view awareness. We believe that our All-Angles Bench offers valuable insights and contribute to bridging the gap between MLLMs and human-level multi-view understanding. The project and benchmark are publicly available at https://danielchyeh.github.io/All-Angles-Bench/.

Large Language Models (LLMs) have shown impressive capabilities in transforming natural language questions about relational databases into SQL queries. Despite recent improvements, small LLMs struggle to handle questions involving multiple tables and complex SQL patterns under a Zero-Shot Learning (ZSL) setting. Supervised Fine-Tuning (SFT) partially compensates for the knowledge deficits in pretrained models but falls short while dealing with queries involving multi-hop reasoning. To bridge this gap, different LLM training strategies to reinforce reasoning capabilities have been proposed, ranging from leveraging a thinking process within ZSL, including reasoning traces in SFT, or adopt Reinforcement Learning (RL) strategies. However, the influence of reasoning on Text2SQL performance is still largely unexplored. This paper investigates to what extent LLM reasoning capabilities influence their Text2SQL performance on four benchmark datasets. To this end, it considers the following LLM settings: (1) ZSL, including general-purpose reasoning or not; (2) SFT, with and without task-specific reasoning traces; (3) RL, exploring the use of different rewarding functions, both the established EXecution accuracy (EX) and a mix with fine-grained ones that also account the precision, recall, and cardinality of partially correct answers; (4) SFT+RL, i.e, a two-stage approach that combines SFT and RL. The results show that general-purpose reasoning under ZSL proves to be ineffective in tackling complex Text2SQL cases. Small LLMs benefit from SFT with reasoning much more than larger ones. RL is generally beneficial across all tested models and datasets. The use of the fine-grained metrics turns out to be the most effective RL strategy. Thanks to RL and the novel text2SQL rewards, the 7B Qwen-Coder-2.5 model performs on par with 400+ Billion ones (including gpt-4o) on the Bird dataset.

Recently, overconfidence in large language models (LLMs) has garnered considerable attention due to its fundamental importance in quantifying the trustworthiness of LLM generation. However, existing approaches prompt the \textit{black box LLMs} to produce their confidence (\textit{verbalized confidence}), which can be subject to many biases and hallucinations. Inspired by a different aspect of overconfidence in cognitive science called \textit{overprecision}, we designed a framework for its study in black box LLMs. This framework contains three main phases: 1) generation, 2) refinement and 3) evaluation. In the generation phase we prompt the LLM to generate answers to numerical questions in the form of intervals with a certain level of confidence. This confidence level is imposed in the prompt and not required for the LLM to generate as in previous approaches. We use various prompting techniques and use the same prompt multiple times to gauge the effects of randomness in the generation process. In the refinement phase, answers from the previous phase are refined to generate better answers. The LLM answers are evaluated and studied in the evaluation phase to understand its internal workings. This study allowed us to gain various insights into LLM overprecision: 1) LLMs are highly uncalibrated for numerical tasks 2) there is no correlation between the length of the interval and the imposed confidence level, which can be symptomatic of a a) lack of understanding of the concept of confidence or b) inability to adjust self-confidence by following instructions, {3) LLM numerical precision differs depending on the task, scale of answer and prompting technique 4) Refinement of answers doesn't improve precision in most cases. We believe this study offers new perspectives on LLM overconfidence and serves as a strong baseline for overprecision in LLMs.

We present the Mu-SHROOM shared task which is focused on detecting hallucinations and other overgeneration mistakes in the output of instruction-tuned large language models (LLMs). Mu-SHROOM addresses general-purpose LLMs in 14 languages, and frames the hallucination detection problem as a span-labeling task. We received 2,618 submissions from 43 participating teams employing diverse methodologies. The large number of submissions underscores the interest of the community in hallucination detection. We present the results of the participating systems and conduct an empirical analysis to identify key factors contributing to strong performance in this task. We also emphasize relevant current challenges, notably the varying degree of hallucinations across languages and the high annotator disagreement when labeling hallucination spans.