Abstract--While large language models (LLMs) are powerful assistants in programming tasks, they may also produce mali cious code. T esting LLM-generated code therefore poses significa nt risks to assessment infrastructure tasked with executing u n-trusted code. T o address these risks, this work focuses on evaluating the security and confidentiality properties of t est environments, reducing the risk that LLM-generated code ma y compromise the assessment infrastructure. We introduce Sa nd-boxEval, a test suite featuring manually crafted test cases that simulate real-world safety scenarios for LLM assessment en vi-ronments in the context of untrusted code execution. The sui te evaluates vulnerabilities to sensitive information expos ure, filesys-tem manipulation, external communication, and other poten tially dangerous operations in the course of assessment activity. We demonstrate the utility of SandboxEval by deploying it on an open-source implementation of Dyff, an established AI asse ssment framework used to evaluate the safety of LLMs at scale. We sho w, first, that the test suite accurately describes limitations placed on an LLM operating under instructions to generate malicious c ode. Second, we show that the test results provide valuable insig hts for developers seeking to harden assessment infrastructur e and identify risks associated with LLM execution activities. There is growing interest in using large language models (LLMs) to assist with code generation due to their ability to produce relevant code for various programming tasks [1, 2, 3 ]. However, using code generated by LLMs involves certain risks, as it may contain subtle bugs or security flaws that are not immediately apparent [4, 5, 6, 7]. For instance, a malicious model developer may intentionally train poisone d LLMs to inject malicious code, subtly manipulating completions to benefit themselves [8, 9, 10]. Through prompt injection, a malicious user may manipulate input to produce harmful outputs [11, 12].

Language models (LMs) have become a staple of the code-writing toolbox. Their pre-training recipe has, however, remained stagnant over recent years, barring the occasional changes in data sourcing and filtering strategies. In particular, research exploring modifications to Code-LMs' pre-training objectives, geared towards improving data efficiency and better disentangling between syntax and semantics, has been noticeably sparse, especially compared with corresponding efforts in natural language LMs. In this work, we examine grounding on obfuscated code as a means of helping Code-LMs look beyond the surface-form syntax and enhance their pre-training sample efficiency. To this end, we compile ObscuraX, a dataset of approximately 55M source and obfuscated code pairs in seven languages. Subsequently, we pre-train ObscuraCoder models, ranging in size from 255M to 2.8B parameters, on a 272B-token corpus that includes ObscuraX and demonstrate that our obfuscation-based pre-training recipe leads to consistent improvements in Code-LMs' abilities compared to both vanilla autoregressive pre-training as well as existing de-obfuscation (DOBF) objectives. ObscuraCoder demonstrates sizeable gains across multiple tests of syntactic and semantic code understanding, along with improved capabilities in multilingual code completion, multilingual code commit summarization, and multi-purpose library-oriented code generation.

Composed Image Retrieval (CIR) facilitates image retrieval through a multimodal query consisting of a reference image and modification text. The reference image defines the retrieval context, while the modification text specifies desired alterations. However, existing CIR datasets predominantly employ coarse-grained modification text (CoarseMT), which inadequately captures fine-grained retrieval intents. This limitation introduces two key challenges: (1) ignoring detailed differences leads to imprecise positive samples, and (2) greater ambiguity arises when retrieving visually similar images. These issues degrade retrieval accuracy, necessitating manual result filtering or repeated queries. To address these limitations, we develop a robust fine-grained CIR data annotation pipeline that minimizes imprecise positive samples and enhances CIR systems' ability to discern modification intents accurately. Using this pipeline, we refine the FashionIQ and CIRR datasets to create two fine-grained CIR datasets: Fine-FashionIQ and Fine-CIRR. Furthermore, we introduce FineCIR, the first CIR framework explicitly designed to parse the modification text. FineCIR effectively captures fine-grained modification semantics and aligns them with ambiguous visual entities, enhancing retrieval precision. Extensive experiments demonstrate that FineCIR consistently outperforms state-of-the-art CIR baselines on both fine-grained and traditional CIR benchmark datasets. Our FineCIR code and fine-grained CIR datasets are available at https://github.com/SDU-L/FineCIR.git.

Model merging acquires general capabilities without extra data or training by combining multiple models' parameters. Previous approaches achieve linear mode connectivity by aligning parameters into the same loss basin using permutation invariance. In this paper, we introduce Model Assembly Learning (MAL), a novel paradigm for model merging that iteratively integrates parameters from diverse models in an open-ended model zoo to enhance the base model's capabilities. Unlike previous works that require identical architectures, MAL allows the merging of heterogeneous architectures and selective parameters across layers. Specifically, the base model can incorporate parameters from different layers of multiple pre-trained models. We systematically investigate the conditions and fundamental settings of heterogeneous parameter merging, addressing all possible mismatches in layer widths between the base and target models. Furthermore, we establish key laws and provide practical guidelines for effectively implementing MAL.

In recent years, the use of databases that analyze trends, sentiments or news to make economic projections or create indicators has gained significant popularity, particularly with the Google Trends platform. This article explores the potential of Google search data to develop a new index that improves economic forecasts, with a particular focus on one of the key components of economic activity: private consumption (64\% of GDP in Peru). By selecting and estimating categorized variables, machine learning techniques are applied, demonstrating that Google data can identify patterns to generate a leading indicator in real time and improve the accuracy of forecasts. Finally, the results show that Google's "Food" and "Tourism" categories significantly reduce projection errors, highlighting the importance of using this information in a segmented manner to improve macroeconomic forecasts.

Generative artificial intelligence (GenAI) or large language models (LLMs) have the potential to revolutionize computational social science, particularly in automated textual analysis. In this paper, we conduct a systematic evaluation of the promises and risks of using LLMs for diverse coding tasks, with social movement studies serving as a case example. We propose a framework for social scientists to incorporate LLMs into text annotation, either as the primary coding decision-maker or as a coding assistant. This framework provides tools for researchers to develop the optimal prompt, and to examine and report the validity and reliability of LLMs as a methodological tool. Additionally, we discuss the associated epistemic risks related to validity, reliability, replicability, and transparency. We conclude with several practical guidelines for using LLMs in text annotation tasks, and how we can better communicate the epistemic risks in research.

Objective: We aim to investigate long-term robotic surgical skill acquisition among surgical residents and the effects of training intervals and fatigue on performance. Methods: For six months, surgical residents participated in three training sessions once a month, surrounding a single 26-hour hospital shift. In each shift, they participated in training sessions scheduled before, during, and after the shift. In each training session, they performed three dry-lab training tasks: Ring Tower Transfer, Knot-Tying, and Suturing. We collected a comprehensive dataset, including videos synchronized with kinematic data, activity tracking, and scans of the suturing pads. Results: We collected a dataset of 972 trials performed by 18 residents of different surgical specializations. Participants demonstrated consistent performance improvement across all tasks. In addition, we found variations in between-shift learning and forgetting across metrics and tasks, and hints for possible effects of fatigue. Conclusion: The findings from our first analysis shed light on the long-term learning processes of robotic surgical skills with extended intervals and varying levels of fatigue. Significance: This study lays the groundwork for future research aimed at optimizing training protocols and enhancing AI applications in surgery, ultimately contributing to improved patient outcomes. The dataset will be made available upon acceptance of our journal submission.

Recent Large Reasoning Models (LRMs), such as DeepSeek-R1 and OpenAI o1, have demonstrated strong performance gains by scaling up the length of Chain-of-Thought (CoT) reasoning during inference. However, a growing concern lies in their tendency to produce excessively long reasoning traces, which are often filled with redundant content (e.g., repeated definitions), over-analysis of simple problems, and superficial exploration of multiple reasoning paths for harder tasks. This inefficiency introduces significant challenges for training, inference, and real-world deployment (e.g., in agent-based systems), where token economy is critical. In this survey, we provide a comprehensive overview of recent efforts aimed at improving reasoning efficiency in LRMs, with a particular focus on the unique challenges that arise in this new paradigm. We identify common patterns of inefficiency, examine methods proposed across the LRM lifecycle, i.e., from pretraining to inference, and discuss promising future directions for research. To support ongoing development, we also maintain a real-time GitHub repository tracking recent progress in the field. We hope this survey serves as a foundation for further exploration and inspires innovation in this rapidly evolving area.

Imitation learning is a popular method for teaching robots new behaviors. However, most existing methods focus on teaching short, isolated skills rather than long, multi-step tasks. To bridge this gap, imitation learning algorithms must not only learn individual skills but also an abstract understanding of how to sequence these skills to perform extended tasks effectively. This paper addresses this challenge by proposing a neuro-symbolic imitation learning framework. Using task demonstrations, the system first learns a symbolic representation that abstracts the low-level state-action space. The learned representation decomposes a task into easier subtasks and allows the system to leverage symbolic planning to generate abstract plans. Subsequently, the system utilizes this task decomposition to learn a set of neural skills capable of refining abstract plans into actionable robot commands. Experimental results in three simulated robotic environments demonstrate that, compared to baselines, our neuro-symbolic approach increases data efficiency, improves generalization capabilities, and facilitates interpretability.

Recent advancements in unmanned aerial vehicle (UAV) technology have opened new avenues for dynamic data collection in challenging environments, such as sports fields during fast-paced sports action. For the purposes of monitoring sport events for dangerous injuries, we envision a coordinated UAV fleet designed to capture high-quality, multi-view video footage of collision events in real-time. The extracted video data is crucial for analyzing athletes' motions and investigating the probability of sports-related traumatic brain injuries (TBI) during impacts. This research implemented a UAV fleet system on the NetLogo platform, utilizing custom collision detection algorithms to compare against traditional TV-coverage strategies. Our system supports decentralized data capture and autonomous processing, providing resilience in the rapidly evolving dynamics of sports collisions. The collaboration algorithm integrates both shared and local data to generate multi-step analyses aimed at determining the efficacy of custom methods in enhancing the accuracy of TBI prediction models. Missions are simulated in real-time within a two-dimensional model, focusing on the strategic capture of collision events that could lead to TBI, while considering operational constraints such as rapid UAV maneuvering and optimal positioning. Preliminary results from the NetLogo simulations suggest that custom collision detection methods offer superior performance over standard TV-coverage strategies by enabling more precise and timely data capture. This comparative analysis highlights the advantages of tailored algorithmic approaches in critical sports safety applications.