As large language models (LLMs) become increasingly capable and widely adopted, benchmarks play a central role in assessing their practical utility. For example, SWE-Bench Verified has emerged as a critical benchmark for evaluating LLMs' software engineering abilities, particularly their aptitude for resolving real-world GitHub issues. Recent LLMs show impressive performance on SWE-Bench, leading to optimism about their capacity for complex coding tasks. However, current evaluation protocols may overstate these models' true capabilities. It is crucial to distinguish LLMs' generalizable problem-solving ability and other learned artifacts. In this work, we introduce two diagnostic tasks: file path identification from issue descriptions alone and ground truth function reproduction with only the current file context and issue description to probe models' underlying knowledge. We present empirical evidence that performance gains on SWE-Bench-Verified may be partially driven by memorization rather than genuine problem-solving. We show that state-of-the-art models achieve up to 76% accuracy in identifying buggy file paths using only issue descriptions, without access to repository structure. This performance is merely up to 53% on tasks from repositories not included in SWE-Bench, pointing to possible data contamination or memorization. Similar patterns are also observed for the function reproduction task, where the verbatim similarity is much higher on SWE-Bench Verified than on other similar coding benchmarks (up to 35% consecutive 5-gram accuracy on SWE-Bench Verified and Full, but only up to 18% for tasks in other benchmarks). These findings raise concerns about the validity of existing results and underscore the need for more robust, contamination-resistant benchmarks to reliably evaluate LLMs' coding abilities.

Modern life has witnessed the explosion of mobile devices. However, besides the valuable features that bring convenience to end users, security and privacy risks still threaten users of mobile apps. The increasing sophistication of these threats in recent years has underscored the need for more advanced and efficient detection approaches. In this chapter, we explore the application of Large Language Models (LLMs) to identify security risks and privacy violations and mitigate them for the mobile application ecosystem. By introducing state-of-the-art research that applied LLMs to mitigate the top 10 common security risks of smartphone platforms, we highlight the feasibility and potential of LLMs to replace traditional analysis methods, such as dynamic and hybrid analysis of mobile apps. As a representative example of LLM-based solutions, we present an approach to detect sensitive data leakage when users share images online, a common behavior of smartphone users nowadays. Finally, we discuss open research challenges.

Variational Autoencoders (VAEs) remain a cornerstone of generative computer vision, yet their training is often plagued by artifacts that degrade reconstruction and generation quality. This paper introduces VIVAT, a systematic approach to mitigating common artifacts in KL-VAE training without requiring radical architectural changes. We present a detailed taxonomy of five prevalent artifacts - color shift, grid patterns, blur, corner and droplet artifacts - and analyze their root causes. Through straightforward modifications, including adjustments to loss weights, padding strategies, and the integration of Spatially Conditional Normalization, we demonstrate significant improvements in VAE performance. Our method achieves state-of-the-art results in image reconstruction metrics (PSNR and SSIM) across multiple benchmarks and enhances text-to-image generation quality, as evidenced by superior CLIP scores. By preserving the simplicity of the KL-VAE framework while addressing its practical challenges, VIVAT offers actionable insights for researchers and practitioners aiming to optimize VAE training.

Understanding the behavior of deep reinforcement learning (DRL) agents -particularly as task and agent sophistication increase- requires more than simple comparison of reward curves, yet standard methods for behavioral analysis remain underdeveloped in DRL. We apply tools from neuroscience and ethology to study DRL agents in a novel, complex, partially observable environment, ForageWorld, designed to capture key aspects of real-world animal foraging- including sparse, depleting resource patches, predator threats, and spatially extended arenas. We use this environment as a platform for applying joint behavioral and neural analysis to agents, revealing detailed, quantitatively grounded insights into agent strategies, memory, and planning. Contrary to common assumptions, we find that model-free RNN-based DRL agents can exhibit structured, planning-like behavior purely through emergent dynamics- without requiring explicit memory modules or world models. Our results show that studying DRL agents like animals -analyzing them with neuroethology-inspired tools that reveal structure in both behavior and neural dynamics- uncovers rich structure in their learning dynamics that would otherwise remain invisible. We distill these tools into a general analysis framework linking core behavioral and representational features to diagnostic methods, which can be reused for a wide range of tasks and agents. As agents grow more complex and autonomous, bridging neuroscience, cognitive science, and AI will be essential- not just for understanding their behavior, but for ensuring safe alignment and maximizing desirable behaviors that are hard to measure via reward. We show how this can be done by drawing on lessons from how biological intelligence is studied.

Current large language model (LLM) agents lack authentic human psychological processes necessary for genuine digital twins and social AI applications. To address this limitation, we present a computational implementation of Global Workspace Theory (GNWT) that integrates human cognitive architecture principles into LLM agents, creating specialized sub-agents for emotion, memory, social norms, planning, and goal-tracking coordinated through a global workspace mechanism. However, authentic digital twins require accurate personality initialization. We therefore develop a novel adventure-based personality test that evaluates true personality through behavioral choices within interactive scenarios, bypassing self-presentation bias found in traditional assessments. Building on these innovations, our CogniPair platform enables digital twins to engage in realistic simulated dating interactions and job interviews before real encounters, providing bidirectional cultural fit assessment for both romantic compatibility and workplace matching. Validation using 551 GNWT-Agents and Columbia University Speed Dating dataset demonstrates 72% correlation with human attraction patterns, 77.8% match prediction accuracy, and 74% agreement in human validation studies. This work advances psychological authenticity in LLM agents and establishes a foundation for intelligent dating platforms and HR technology solutions.

Neural models learn representations of high-dimensional data on low-dimensional manifolds. Multiple factors, including stochasticities in the training process, model architectures, and additional inductive biases, may induce different representations, even when learning the same task on the same data. However, it has recently been shown that when a latent structure is shared between distinct latent spaces, relative distances between representations can be preserved, up to distortions. Building on this idea, we demonstrate that exploiting the differential-geometric structure of latent spaces of neural models, it is possible to capture precisely the transformations between representational spaces trained on similar data distributions. Specifically, we assume that distinct neural models parametrize approximately the same underlying manifold, and introduce a representation based on the pullback metric that captures the intrinsic structure of the latent space, while scaling efficiently to large models. We validate experimentally our method on model stitching and retrieval tasks, covering autoencoders and vision foundation discriminative models, across diverse architectures, datasets, and pretraining schemes.

Instruction tuning is essential for Large Language Models (LLMs) to effectively follow user instructions. To improve training efficiency and reduce data redundancy, recent works use LLM-based scoring functions, e.g., Instruction-Following Difficulty (IFD), to select high-quality instruction-tuning data with scores above a threshold. While these data selection methods often lead to models that can match or even exceed the performance of models trained on the full datasets, we identify two key limitations: (i) they assess quality at the sample level, ignoring token-level informativeness; and (ii) they overlook the robustness of the scoring method, often selecting a sample due to superficial lexical features instead of its true quality. In this work, we propose Token-Selective HIeRarchical Data Selection for Instruction Tuning (T-SHIRT), a novel data selection framework that introduces a new scoring method to include only informative tokens in quality evaluation and also promotes robust and reliable samples whose neighbors also show high quality with less local inconsistencies. We demonstrate that models instruction-tuned on a curated dataset (only 5% of the original size) using T-SHIRT can outperform those trained on the entire large-scale dataset by up to 5.48 points on average across eight benchmarks. Across various LLMs and training set scales, our method consistently surpasses existing state-of-the-art data selection techniques, while also remaining both cost-effective and highly efficient. For instance, by using GPT-2 for score computation, we are able to process a dataset of 52k samples in 40 minutes on a single GPU. Our code is available at https://github.com/Dynamite321/T-SHIRT.

Background and Objective: Plane reformatting for four-dimensional phase contrast MRI (4D flow MRI) is time-consuming and prone to inter-observer variability, which limits fast cardiovascular flow assessment. Deep reinforcement learning (DRL) trains agents to iteratively adjust plane position and orientation, enabling accurate plane reformatting without the need for detailed landmarks, making it suitable for images with limited contrast and resolution such as 4D flow MRI. However, current DRL methods assume that test volumes share the same spatial alignment as the training data, limiting generalization across scanners and institutions. To address this limitation, we introduce AdaPR (Adaptive Plane Reformatting), a DRL framework that uses a local coordinate system to navigate volumes with arbitrary positions and orientations. Methods: We implemented AdaPR using the Asynchronous Advantage Actor-Critic (A3C) algorithm and validated it on 88 4D flow MRI datasets acquired from multiple vendors, including patients with congenital heart disease. Results: AdaPR achieved a mean angular error of 6.32 +/- 4.15 degrees and a distance error of 3.40 +/- 2.75 mm, outperforming global-coordinate DRL methods and alternative non-DRL methods. AdaPR maintained consistent accuracy under different volume orientations and positions. Flow measurements from AdaPR planes showed no significant differences compared to two manual observers, with excellent correlation (R^2 = 0.972 and R^2 = 0.968), comparable to inter-observer agreement (R^2 = 0.969). Conclusion: AdaPR provides robust, orientation-independent plane reformatting for 4D flow MRI, achieving flow quantification comparable to expert observers. Its adaptability across datasets and scanners makes it a promising candidate for medical imaging applications beyond 4D flow MRI.

The full statistics of all the criteria grouped by the panel assignments are presented in Table 7. Tables 8 and A.1 represent the statistics of the generated scores and rationales for criteria annotation. As we can see, the distributions of criterion-based scores for most criteria are largely comparable between expert-written and synthetic datasets, despite the underlying evaluated instruction-answer pairs being entirely distinct and non-overlapping. This is particularly evident in the mean, standard deviation, and mode of scores, which, across a wide range of criteria types, demonstrate close alignment - suggesting that criterion-level assessment remains consistent across both data sources. Tables 8 and A.1 suggest that synthetically generated texts (both instructions and rationales) are lengthier, being at the same time less original than those written by the experts. Tables also show that DeepSeek-R1 tends to assign a mediocre score of 1 rather than choosing extreme values. Despite these statistical and stylistic differences in commentary, the synthetic dataset remains a viable resource for training the LLM-as-a-Judge Family, especially considering the overall similarity in criterion-based scores. Thus, while the expert-written feedback exhibits optimized brevity and contextual appropriateness, the synthetic commentary maintains an adequate level of informative-ness and coherence.

Large language models (LLMs) memorize a vast amount of prior knowledge from the Internet that helps them on downstream tasks but also may notoriously sway their outputs towards wrong or biased answers. In this work, we test how the knowledge about popular subjects hurt the accuracy of vision language models (VLMs) on standard, objective visual tasks of counting and identification. We find that state-of-the-art VLMs are strongly biased (e.g., unable to recognize the 4th stripe has been added to a 3-stripe Adidas logo) scoring an average of 17.05% accuracy in counting (e.g., counting stripes in an Adidas-like logo) across 7 diverse domains from animals, logos, chess, board games, optical illusions, to patterned grids. Removing image backgrounds nearly doubles accuracy (21.09 percentage points), revealing that contextual visual cues trigger these biased responses. Further analysis of VLMs' reasoning patterns shows that counting accuracy initially rises with thinking tokens, reaching ~40%, before declining with excessive reasoning. Our work presents an interesting failure mode in VLMs and a human-supervised automated framework for testing VLM biases. Code and data are available at: vlmsarebiased.github.io.