T propose o address a no the vel challenge Agent-as-a-Judge of evaluating framew time-v ork. Our arying method and construct complex s answers, task-specific we judg answer of ten e a frontier g correctness ents based agentic and on a search source tree-structured systems attribution.

Active testing enables label-efficient evaluation of predictive models through careful data acquisition, but it can pose a significant computational cost. We identify cost-saving measures that enable active testing to be scaled up to large language models (LLMs). In particular we show that the surrogate model used to guide data acquisition can be constructed cheaply using in-context learning, does not require updating within an active-testing loop, and can be smaller than the target model. We even find we can make good data-acquisition decisions without making predictions with the target model. As a result we are able to achieve much more accurate evaluations of LLM performance relative to using randomly acquired data. We additionally introduce a bootstrap estimator of evaluation error, which we show to be a useful indicator of how well active testing is working within a single run.

AI systems are increasingly able to autonomously conduct realistic software engineering tasks, and may soon be deployed to automate machine learning (ML) R&D itself. Frontier AI systems may be deployed in safety-critical settings, including to help ensure the safety of future systems. Unfortunately, frontier and future systems may not be sufficiently trustworthy, and there is evidence that these systems may even be misaligned with their developers or users. Therefore, we investigate the capabilities of AI agents to act against the interests of their users when conducting ML engineering, by sabotaging ML models, sandbagging their performance, and subverting oversight mechanisms. First, we extend MLE-Bench, a benchmark for realistic ML tasks, with code-sabotage tasks such as implanting backdoors and purposefully causing generalisation failures.

As distributed optimization scales to meet the demands of Large Language Model (LLM) training, hardware failures become increasingly non-negligible. Existing fault-tolerant training methods often introduce significant computational or memory overhead, demanding additional resources. To address this challenge, we propose Memory-and Computation-efficient Fault-tolerant Optimization (MeCeFO), a novel algorithm that ensures robust training with minimal overhead. When a computing node fails, MeCeFO seamlessly transfers its training task to a neighboring node while employing memory-and computation-efficient algorithmic optimizations to minimize the extra workload imposed on the neighboring node handling both tasks. MeCeFO leverages three key algorithmic designs: (i) Skip-connection, which drops the multi-head attention (MHA) module during backpropagation for memory-and computation-efficient approximation; (ii) Recomputation, which reduces activation memory in feedforward networks (FFNs); and (iii) Low-rank gradient approximation, enabling efficient estimation of FFN weight matrix gradients. Theoretically, MeCeFO matches the convergence rate of conventional distributed training, with a rate of O(1/ nT), where n is the data parallelism size and T is the number of iterations. Empirically, MeCeFO maintains robust performance under high failure rates, incurring only a 4.18% drop in throughput, demonstrating 5.0 to 6.7 greater resilience than previous SOTA approaches.

Autoregressive models have driven remarkable progress in language modeling. Their foundational reliance on discrete tokens, unidirectional context, and singlepass decoding, while central to their success, also inspires the exploration of a design space that could offer new axes of modeling flexibility. In this work, we explore an alternative paradigm, shifting language modeling from a discrete token space to a continuous latent space. We propose a novel framework TarFlowLM, that employs transformer-based autoregressive normalizing flows [73] to model these continuous representations. This approach unlocks substantial flexibility, enabling the construction of models that can capture global bi-directional context through stacked, alternating-direction autoregressive transformations, support block-wise generation with flexible token patch sizes, and facilitate a hierarchical multi-pass generation process. We further propose new mixture-based coupling transformations designed to capture complex dependencies within the latent space shaped by discrete data, and demonstrate theoretical connections to conventional discrete autoregressive models. Extensive experiments on language modeling benchmarks demonstrate strong likelihood performance and highlight the flexible modeling capabilities inherent in our framework.

Large language models (LLMs) have shown impressive capabilities across tasks such as mathematics, coding, and reasoning, yet their learning ability, which is crucial for adapting to dynamic environments and acquiring new knowledge, remains underexplored. In this work, we address this gap by introducing a framework inspired by cognitive psychology and education. Specifically, we decompose general learning ability into three distinct, complementary dimensions: Learning from Instructor (acquiring knowledge via explicit guidance), Learning from Concept (internalizing abstract structures and generalizing to new contexts), and Learning from Experience (adapting through accumulated exploration and feedback). We conduct a comprehensive empirical study across the three learning dimensions and identify several insightful findings, such as (i) interaction improves learning; (ii) conceptual understanding is scale-emergent and benefits larger models; and (iii) LLMs are effective few-shot learners but not many-shot learners. Based on our framework and empirical findings, we introduce a benchmark that provides a unified and realistic evaluation of LLMs' general learning abilities across three learning cognition dimensions. It enables diagnostic insights and supports evaluation and development of more adaptive and human-like models. The code is available here 2.

Indoor environments are the primary setting where humans spend most of their daily lives. Yet, computationally creating digital twins of these 3D spaces from captured images remains challenging. Factors such as the difficulty of accurate camera pose estimation from indoor images [28, 11, 1] and structural distortions in the resulting 3D reconstructions [22, 12, 21] hinder the development of robust, accurate, and user-friendly solutions for replicating indoor scenes in the digital world. As indoor scenes are typically rich in planar structures such as floors, ceilings, and walls, as well as planar furniture like tables and cabinets, planar primitives are well-suited representations for the accurate 3D reconstruction of indoor scenes. As a result, there has been significant interest among the research community in planar 3D reconstruction in recent years. Planar reconstruction approaches include feedforward solutions in monocular [40, 16, 27, 24, 18, 42] and two-view [11, 1, 28] settings, and per-scene optimization approaches [29, 38, 3, 9] that leverage posed multi-view inputs with the assistance of the feedforward models were studied. However, these approaches face two key limitations: Annotation dependence for feedforward methods: Learning feedforward models [36, 24, 28] typically requires accurate plane masks and 3D plane annotations from monocular or binocular inputs.

A major challenge in training VLM agents, compared to LLM agents, is that states shift from simple texts to complex visual observations, which introduces partial observability and demands robust world modeling. We ask: can VLM agents build internal world models through explicit visual state reasoning? In this work, we architecturally enforce and reward VLM agent's reasoning process via reinforcement learning (RL), formulating the problem as a Partially Observable Markov Decision Process (POMDP). We demonstrate that structuring agent's reasoning into StateEstimation("what is the current state?") and TransitionModeling ("what is next?") is critical by studying five reasoning strategies. Investigating how agents should ground visual states and represent these internal beliefs, we reveal the optimal representations are task-dependent: Natural Language excels at capturing semantic relationships for general tasks, while Structured formats are essential for high-precision manipulation. These insights motivate our approach to reward shaping and credit assignment. We leverage a WorldModelingReward to densely rewards the agent's turn-by-turn state predictions, while our Bi-Level General Advantage Estimation (Bi-Level GAE) enables turn-aware credit assignment. Through such world model reasoning, we enable a 3B model to achieve performance of 0.82 on a set of five diverse agent tasks, nearly 3 improvement over its untrained counterpart (0.21) and surpassing proprietary reasoning models like GPT-5 (0.75), Gemini 2.5 Pro (0.67) and Claude 4.5 (0.62). All experiments are supported by our VAGEN framework, a scalable system for training and analyzing multi-turn VLM agents across diverse visual environments.

We introduce a full-stack framework that scales up reasoning in vision-language models (VLMs) to long videos, leveraging reinforcement learning. We address the unique challenges of long video reasoning by integrating three critical components: (1) a large-scale dataset, LongVideo-Reason, comprising 104K long video QA pairs with high-quality reasoning annotations across diverse domains such as sports, games, and vlogs; (2) a two-stage training pipeline that extends VLMs with chainof-thought supervised fine-tuning (CoT-SFT) and reinforcement learning (RL); and (3) a training infrastructure for long video RL, named Multi-modal Reinforcement Sequence Parallelism (MR-SP), which incorporates sequence parallelism and a vLLM-based engine tailored for long video, using cached video embeddings for efficient rollout and prefilling. In our experiments, LongVILA-R1-7B achieves strong performance on video benchmarks, reaching 65.1% and 71.1% accuracy on VideoMME without and with subtitles, respectively, and consistently outperforming LongVILA-7B across multiple benchmarks. Moreover, LongVILA-R1-7B supports processing up to 8,192 video frames per video, and configurable FPS settings. Notably, our MR-SP system achieves up to 2.1 speedup on long video RL training. In addition, we release our training system for public availability that supports RL training on various modalities (video, text, and audio), various models (VILA and Qwen series), and even image and video generation models. On a single A100 node (8 GPUs), it supports RL training on hour-long videos (e.g., 3,600 frames). Code and models are available at https://github.com/NVlabs/Long-RL

With the significant progress of large reasoning models in complex coding and reasoning tasks, existing benchmarks, like LiveCodeBench and CodeElo, are insufficient to evaluate the coding capabilities of large language models (LLMs) in real competition environments. Moreover, current evaluation metrics such as Pass@K fail to capture the reflective abilities of reasoning models. To address these challenges, we propose ICPC-Eval, a top-level competitive coding benchmark designed to probing the frontiers of LLM reasoning. ICPC-Eval includes 118 carefully curated problems from 11 recent ICPC contests held in various regions of the world, offering three key contributions: 1) A challenging realistic ICPC competition scenario, featuring a problem type and difficulty distribution consistent with actual contests.