This paper introduces UI-TARS, a native GUI agent model that solely perceives the screenshots as input and performs human-like interactions (e.g., keyboard and mouse operations). Unlike prevailing agent frameworks that depend on heavily wrapped commercial models (e.g., GPT-4o) with expert-crafted prompts and workflows, UI-TARS is an end-to-end model that outperforms these sophisticated frameworks. Experiments demonstrate its superior performance: UI-TARS achieves SOTA performance in 10+ GUI agent benchmarks evaluating perception, grounding, and GUI task execution (see below). Notably, in the OSWorld benchmark, UI-TARS achieves scores of 24.6 with 50 steps and 22.7 with 15 steps, outperforming Claude's 22.0 and 14.9 respectively. In AndroidWorld, UI-TARS achieves 46.6, surpassing GPT-4o's 34.5. UI-TARS incorporates several key innovations: (1) Enhanced Perception: leveraging a large-scale dataset of GUI screenshots for context-aware understanding of UI elements and precise captioning; (2) Unified Action Modeling, which standardizes actions into a unified space across platforms and achieves precise grounding and interaction through large-scale action traces; (3) System-2 Reasoning, which incorporates deliberate reasoning into multi-step decision making, involving multiple reasoning patterns such as task decomposition, reflection thinking, milestone recognition, etc. (4) Iterative Training with Reflective Online Traces, which addresses the data bottleneck by automatically collecting, filtering, and reflectively refining new interaction traces on hundreds of virtual machines. Through iterative training and reflection tuning, UI-TARS continuously learns from its mistakes and adapts to unforeseen situations with minimal human intervention. We also analyze the evolution path of GUI agents to guide the further development of this domain.

From my very personal point of view, the lifelong learning paradigm is a vague concept and is sometimes evoked for studying different scenarios that can be named otherwise (like transfer learning). However, I think that the framework studied here (from Balcan et al., 2015) is a very pertinent "lifelong problem". The authors present an honest work in the right direction. The proofs are not trivial but, as I explain below, the contribution appears to me insufficient for NIPS. The risk bound minimized by the learning algorithm may be very high, as it relies on the VC-dimension of the predictors.

Figure 1: Training models to optimize test-time compute and learn "how to discover" correct responses, as opposed to the traditional learning paradigm of learning "what answer" to output. The major strategy to improve large language models (LLMs) thus far has been to use more and more high-quality data for supervised fine-tuning (SFT) or reinforcement learning (RL). Unfortunately, it seems this form of scaling will soon hit a wall, with the scaling laws for pre-training plateauing, and with reports that high-quality text data for training maybe exhausted by 2028, particularly for more difficult tasks, like solving reasoning problems which seems to require scaling current data by about 100x to see any significant improvement. The current performance of LLMs on problems from these hard tasks remains underwhelming (see example). There is thus a pressing need for data-efficient methods for training LLMs that extend beyond data scaling and can address more complex challenges.

The authors present an algorithm for lifelong representation learning that adapts variational autoencoders to the lifelong learning setting. The framework is presented as a full generative process, where a set of latent factors are (selectively) shared across tasks, and the tasks themselves are generated by an unknown distribution. The algorithm optimizes for the reconstruction error with a regularization based on the MDL principle that has been studied for learning disentangled representations. The algorithm automatically detects distribution shifts (i.e., task changes) and avoids catastrophic forgetting by "hallucinating" data for previous tasks while training on a new one. The authors show empirically that their algorithm is able to extract relevant semantic knowledge from one task and transfer it to the next.

This paper investigates the role of AI assistants, specifically OpenAI's ChatGPT, in teaching formal methods (FM) to undergraduate students, using the B-method as a formal specification technique. While existing studies demonstrate the effectiveness of AI in coding tasks, no study reports on its impact on formal specifications. We examine whether ChatGPT provides an advantage when writing B-specifications and analyse student trust in its outputs. Our findings indicate that the AI does not help students to enhance the correctness of their specifications, with low trust correlating to better outcomes. Additionally, we identify a behavioural pattern with which to interact with ChatGPT which may influence the correctness of B-specifications.

The integration of AI in education offers significant potential to enhance learning efficiency. Large Language Models (LLMs), such as ChatGPT, Gemini, and Llama, allow students to query a wide range of topics, providing unprecedented flexibility. However, LLMs face challenges, such as handling varying content relevance and lack of personalization. To address these challenges, we propose TutorLLM, a personalized learning recommender LLM system based on Knowledge Tracing (KT) and Retrieval-Augmented Generation (RAG). The novelty of TutorLLM lies in its unique combination of KT and RAG techniques with LLMs, which enables dynamic retrieval of context-specific knowledge and provides personalized learning recommendations based on the student's personal learning state. Specifically, this integration allows TutorLLM to tailor responses based on individual learning states predicted by the Multi-Features with Latent Relations BERT-based KT (MLFBK) model and to enhance response accuracy with a Scraper model. The evaluation includes user assessment questionnaires and performance metrics, demonstrating a 10\% improvement in user satisfaction and a 5\% increase in quiz scores compared to using general LLMs alone.

Procedural activity understanding requires perceiving human actions in terms of a broader task, where multiple keysteps are performed in sequence across a long video to reach a final goal state---such as the steps of a recipe or the steps of a DIY fix-it task. Prior work largely treats keystep recognition in isolation of this broader structure, or else rigidly confines keysteps to align with a particular sequential script. We propose discovering a task graph automatically from how-to videos to represent probabilistically how people tend to execute keysteps, then leverage this graph to regularize keystep recognition in novel videos. On multiple datasets of real-world instructional video, we show the impact: more reliable zero-shot keystep localization and improved video representation learning, exceeding the state of the art.

Recent research on online Gradient Balancing (GraB) has revealed that there exist permutation-based example orderings for SGD that are guaranteed to outperform random reshuffling (RR). Whereas RR arbitrarily permutes training examples, GraB leverages stale gradients from prior epochs to order examples -- achieving a provably faster convergence rate than RR. However, GraB is limited by design: while it demonstrates an impressive ability to scale-up training on centralized data, it does not naturally extend to modern distributed ML workloads. We therefore propose Coordinated Distributed GraB (CD-GraB), which uses insights from prior work on kernel thinning to translate the benefits of provably faster permutation-based example ordering to distributed settings. With negligible overhead, CD-GraB exhibits a linear speedup in convergence rate over centralized GraB and outperforms distributed RR on a variety of benchmark tasks.

Image classification and segmentation are common applications of deep learning to radiology. While many tasks can be framed using either classification or segmentation, classification has historically been cheaper to label and more widely used. However, recent work has drastically reduced the cost of training segmentation networks. First, we use an information theoretic approach to analyze why segmentation vs. classification models may achieve different performance on the same dataset and overarching task. We use our analysis and experiments to summarize the benefits of switching from segmentation to classification, including: improved sample efficiency, enabling improved performance with fewer labeled images (up to an order of magnitude lower), on low-prevalence classes, and on certain rare subgroups (up to 161.1\% improved recall); improved robustness to spurious correlations (up to 44.8\% improved robust AUROC); and improved model interpretability, evaluation, and error analysis.

Offline-to-online reinforcement learning (RL) is a training paradigm that combines pre-training on a pre-collected dataset with fine-tuning in an online environment. However, the incorporation of online fine-tuning can intensify the well-known distributional shift problem. Existing solutions tackle this problem by imposing a policy constraint on the policy improvement objective in both offline and online learning. They typically advocate a single balance between policy improvement and constraints across diverse data collections. This one-size-fits-all manner may not optimally leverage each collected sample due to the significant variation in data quality across different states.