The applications of large language models (LLMs) in various biological domains have been explored recently, but their reasoning ability in complex biological systems, such as pathways, remains underexplored, which is crucial for predicting biological phenomena, formulating hypotheses, and designing experiments. This work explores the potential of LLMs in pathway reasoning. We introduce BioMaze, a dataset with 5.1K complex pathway problems derived from real research, covering various biological contexts including natural dynamic changes, disturbances, additional intervention conditions, and multi-scale research targets. Our evaluation of methods such as CoT and graph-augmented reasoning, shows that LLMs struggle with pathway reasoning, especially in perturbed systems. To address this, we propose PathSeeker, an LLM agent that enhances reasoning through interactive subgraph-based navigation, enabling a more effective approach to handling the complexities of biological systems in a scientifically aligned manner. The dataset and code are available at https://github.com/zhao-ht/BioMaze.

Armed with my shopping list, it types each item into the search bar of a supermarket website, then uses its cursor to click. Watching what appears to be a digital ghost do this usually mundane task is strangely transfixing. "Are you sure it's not just a person in India?" my husband asks, peering over my shoulder. Made available to UK users last month, it has a similar text interface and conversational tone to ChatGPT, but rather than just answering questions, it can actually do things – provided they involve navigating a web browser. Hot on the heels of large language models, AI agents have been trumpeted as the next big thing, and you can see the appeal: a digital assistant that can complete practical tasks is more compelling than one that can just talk back.

How can we build generalist robot systems? Scale may not be enough due to the significant multimodality of robotics tasks, lack of easily accessible data and the challenges of deploying on physical hardware. Meanwhile, most deployed robotic systems today are inherently modular and can leverage the independent generalization capabilities of each module to perform well. Therefore, this thesis seeks to tackle the task of building generalist robot agents by integrating these components into one: combining modularity with large-scale learning for general purpose robot control. The first question we consider is: how can we build modularity and hierarchy into learning systems? Our key insight is that rather than having the agent learn hierarchy and low-level control end-to-end, we can enforce modularity via planning to enable more efficient and capable robot learners. Next, we come to the role of scale in building generalist robot systems. To scale, neural networks require vast amounts of diverse data, expressive architectures to fit the data and a source of supervision to generate the data. We leverage a powerful supervision source: classical planning, which can generalize, but is expensive to run and requires access to privileged information to perform well in practice. We use these planners to supervise large-scale policy learning in simulation to produce generalist agents. Finally, we consider how to unify modularity with large-scale policy learning to build real-world robot systems capable of performing zero-shot manipulation. We do so by tightly integrating key ingredients of modular high and mid-level planning, learned local control, procedural scene generation and large-scale policy learning for sim2real transfer. We demonstrate that this recipe can produce a single, generalist agent that can solve challenging long-horizon manipulation tasks in the real world.

The paper introduces ExKG-LLM, a framework designed to automate the expansion of cognitive neuroscience knowledge graphs (CNKG) using large language models (LLMs). It addresses limitations in existing tools by enhancing accuracy, completeness, and usefulness in CNKG. The framework leverages a large dataset of scientific papers and clinical reports, applying state-of-the-art LLMs to extract, optimize, and integrate new entities and relationships. Evaluation metrics include precision, recall, and graph density. Results show significant improvements: precision (0.80, +6.67%), recall (0.81, +15.71%), F1 score (0.805, +11.81%), and increased edge nodes (21.13% and 31.92%). Graph density slightly decreased, reflecting a broader but more fragmented structure. Engagement rates rose by 20%, while CNKG diameter increased to 15, indicating a more distributed structure. Time complexity improved to O(n log n), but space complexity rose to O(n2), indicating higher memory usage. ExKG-LLM demonstrates potential for enhancing knowledge generation, semantic search, and clinical decision-making in cognitive neuroscience, adaptable to broader scientific fields.

Assessing the safety of vision-language models (VLMs) in autonomous driving is particularly important; however, existing work mainly focuses on traditional benchmark evaluations. As interactive components within autonomous driving systems, VLMs must maintain strong safety cognition during interactions. From this perspective, we propose a novel evaluation method: Safety Cognitive Driving Benchmark (SCD-Bench) . To address the large-scale annotation challenge for SCD-Bench, we develop the Autonomous Driving Image-Text Annotation System (ADA) . Additionally, to ensure data quality in SCD-Bench, our dataset undergoes manual refinement by experts with professional knowledge in autonomous driving. We further develop an automated evaluation method based on large language models (LLMs). To verify its effectiveness, we compare its evaluation results with those of expert human evaluations, achieving a consistency rate of 99.74%. Preliminary experimental results indicate that existing open-source models still lack sufficient safety cognition, showing a significant gap compared to GPT-4o. Notably, lightweight models (1B-4B) demonstrate minimal safety cognition. However, since lightweight models are crucial for autonomous driving systems, this presents a significant challenge for integrating VLMs into the field.

Large vision-language models (LVLMs) have demonstrated remarkable achievements, yet the generation of non-factual responses remains prevalent in fact-seeking question answering (QA). Current multimodal fact-seeking benchmarks primarily focus on comparing model outputs to ground truth answers, providing limited insights into the performance of modality-specific modules. To bridge this gap, we introduce VisualSimpleQA, a multimodal fact-seeking benchmark with two key features. First, it enables streamlined and decoupled evaluation of LVLMs in visual and linguistic modalities. Second, it incorporates well-defined difficulty criteria to guide human annotation and facilitates the extraction of a challenging subset, VisualSimpleQA-hard. Experiments on 15 LVLMs show that even state-of-the-art models such as GPT-4o achieve merely 60%+ correctness in multimodal fact-seeking QA on VisualSimpleQA and 30%+ on VisualSimpleQA-hard. Furthermore, the decoupled evaluation across these models highlights substantial opportunities for improvement in both visual and linguistic modules. The dataset is available at https://huggingface.co/datasets/WYLing/VisualSimpleQA.

W e propose DiffCLIP, a novel vision-language model that extends the differential attention mechanism to CLIP architectures. Differential attention was originally developed for large language models to amplify relevant context while canceling out noisy information. In this work, we integrate this mechanism into CLIP's dual encoder (image and text) framework. With minimal additional parameters, Dif-fCLIP achieves superior performance on image-text understanding tasks. Across zero-shot classification, retrieval, and robustness benchmarks, DiffCLIP consistently outperforms baseline CLIP models. Notably, these gains come with negligible computational overhead, demonstrating that differential attention can significantly enhance multi-modal representations without sacrificing efficiency. Code can be found at https://github.com/hammoudhasan/

We compare the effectiveness of zero-shot Chain-of-Thought (CoT) prompting in Japanese and English using ChatGPT-3.5 and 4o-mini. The technique of zero-shot CoT, which involves appending a phrase such as "Let's think step by step" to a prompt to encourage reasoning before answering, has been shown to offer LLM performance improvements in mathematical and reasoning tasks, particularly in English. We investigate how these effects transfer to Japanese using the Japanese Multi-task Language Understanding Benchmark (JMMLU) and the Multi-task Language Understanding Benchmark (MMLU). Our results show that while zero-shot CoT prompting can lead to notable performance gains for some prompt categories in GPT-3.5, its impact in GPT-4o-mini is associated with significant performance declines. However, for Japanese prompts there remain certain categories, such as college mathematics and abstract algebra, that still exhibit improvements, despite the broader trend of diminishing effectiveness in more advanced models.

Humans can flexibly switch between different modes of thinking based on task complexity: from rapid intuitive judgments to in-depth analytical understanding. However, current Graphical User Interface (GUI) grounding systems which locate interface elements based on natural language instructions rely solely on immediate prediction without reasoning, struggling to understand complex interface layouts with nested structures and hierarchical relationships, limiting their effectiveness on complex interfaces. Inspired by human dual-system cognition, we present Focus, a novel GUI grounding framework that combines fast prediction with systematic analysis. The framework dynamically switches between rapid and deliberate processing through an adaptive system switching based on task complexity, optimizing both efficiency and accuracy. Focus decomposes grounding into progressive stages: interface summarization, visual focused analysis, and precise coordinate prediction. This structured decomposition enables systematic understanding of both interface layouts and visual relationships. Extensive experiments show that Focus achieves state-of-the-art performance using only 300K of the training data with a 2B parameter model compared to existing approaches. Focus demonstrates superior performance particularly in complex GUI scenarios, achieving 77.4% average accuracy on ScreenSpot and 13.3% on the more challenging ScreenSpot-Pro. Our analysis reveals the effectiveness of this dual-system approach while demonstrating its potential for improving complex GUI interaction scenarios.

Large language models (LLMs) have achieved remarkable success in natural language processing, yet their performance and computational costs vary significantly. LLM routers play a crucial role in dynamically balancing these tradeoffs. While previous studies have primarily focused on routing efficiency, security vulnerabilities throughout the entire LLM router life cycle, from training to inference, remain largely unexplored. In this paper, we present a comprehensive investigation into the life-cycle routing vulnerabilities of LLM routers. We evaluate both white-box and black-box adversarial robustness, as well as backdoor robustness, across several representative routing models under extensive experimental settings. Our experiments uncover several key findings: 1) Mainstream DNN-based routers tend to exhibit the weakest adversarial and backdoor robustness, largely due to their strong feature extraction capabilities that amplify vulnerabilities during both training and inference; 2) Training-free routers demonstrate the strongest robustness across different attack types, benefiting from the absence of learnable parameters that can be manipulated. These findings highlight critical security risks spanning the entire life cycle of LLM routers and provide insights for developing more robust models. In recent years, large language models (LLMs) such as GPT-3.5 (Brown et al., 2020), GPT-4 (Achiam et al., 2023), and PaLM 2 (Anil et al., 2023) have achieved significant progress in natural language processing tasks, finding widespread applications in open-domain dialogue, question answering, code generation, and other tasks (Gu, 2023; Zhuang et al., 2023; Ghosh et al., 2024). However, different LLMs vary in terms of training data, model size, and computational cost, leading to differences in their strengths, weaknesses, and overall capabilities. Generally, larger models tend to exhibit stronger performance but come with higher inference costs, whereas smaller models are more computationally efficient but have limited capability in handling complex tasks. LLM Routing (Ding et al., 2024; Ong et al., 2024; Hu et al., 2024) is a state-of-the-art optimization strategy designed to mitigate this trade-off and achieve a balance between response quality and computational cost.