×

Academic Intelligence · Curated Daily

探索全球前沿学术脉络

AcademicHub 汇聚顶级期刊与预印本平台的实时文献。定制您的专属科研雷达,利用大语言模型自动生成交叉领域文献分析简报。

作者: Jie Tan ×
换一批
01.
arXiv (CS.AI) 2026-06-19

Data Standards for Humanoid Robotics: The Missing Infrastructure for Physical AI

arXiv:2606.19769v1 Announce Type: cross Abstract: The scalability of humanoid robots will depend not only on models and hardware, but also on whether physical experience can accumulate across robots, tasks, organizations, and time. Drawing on the authors' work in developing ISO/WD 26264-1, Humanoid robot datasets – Part 1: General requirements, within ISO/TC 299/WG 16, this article argues that data standards are becoming foundational infrastructure for Physical AI. We develop three insights. First, humanoid robot data is embodied interaction data, not a collection of isolated digital samples; a useful dataset must preserve the relationship among robot body, action, task, scene, execution trace, and outcome. Second, its value depends on physical coherence: multimodal streams are reusable only when timing, coordinate frames, calibration, kinematics, units, and synchronization assumptions remain inspectable. Third, the main bottleneck is not only data scarcity, but non-cumulative data caused by high collection costs, data silos, and inconsistent evaluation. We argue that humanoid robot data standards address these bottlenecks by making embodied experience interpretable, shareable, traceable, and reusable. A general standard should provide horizontal infrastructure for lifecycle management, metadata, provenance, quality, versioning, and traceability, while capability-specific parts should define domain grammar for manipulation, locomotion, human-robot interaction, cognition, and future humanoid capabilities. As AI moves from screens into bodies, data standards must evolve from organizing digital information to structuring physical interaction.

02.
arXiv (CS.CL) 2026-06-11

Verifiable Environments Are LEGO Bricks: Recursive Composition for Reasoning Generalization

Reinforcement Learning (RL) with verifiable environments has emerged as a powerful approach for enhancing the reasoning capabilities of Large Language Models (LLMs). While prior research demonstrates that scaling environment quantity improves RL performance, existing manual or individual construction methods suffer from linear scaling limits, thereby hindering scalable reasoning generalization. This paper introduces RACES (Recursive Automated Composition for Environment Scaling), a framework that conceptualizes verifiable environments as composable building blocks that can be recursively assembled. The key insight is that when the codomain (output type) of one environment matches the domain (input type) of another, they can be automatically fused into a new verifiable environment, enabling recursive composition. RACES is implemented with 300 individual environments and defines a set of composition operators (\textsc{SEQUENTIAL}, \textsc{PARALLEL}, \textsc{SORT}, and \textsc{SELECT}) that induce diverse reasoning patterns. Extensive experiments show that RL training on these composite environments consistently enhances reasoning generalization. Specifically, RACES improves DeepSeek-R1-Distill-Qwen-14B by an average of 3.1 points (from 48.2 to 51.3) and boosts Qwen3-14B performance from 58.8 to 61.1 on six benchmarks, which are unseen during the construction of training environments. Moreover, RACES achieves performance comparable to training on 300 individual environments using only 50 base environments, demonstrating significant efficiency in environment utilization.

03.
arXiv (CS.AI) 2026-06-15

FlexMS: A Unified Public Benchmark for Molecule Tandem Mass Spectrum Prediction

arXiv:2602.22822v3 Announce Type: replace Abstract: Tandem mass spectrometry (MS/MS) is central to small molecule identification, but current deep learning systems for spectrum prediction still remain difficult to evaluate and deploy in practice. While novel architectures constantly claim state-of-the-art performance, inconsistent metadata conditioning and entangled preprocessing pipelines hinder fair architectural comparisons. Besides, existing evaluations are often restricted to curated datasets, failing to capture the heterogeneity and cross-domain shifts of real-world metabolomics. Furthermore, current benchmarks lack difficulty-aware diagnostics and leave blind to how models behave under specific compute or data constraints. To address this, we present FlexMS, a modular public-data benchmark framework that standardizes MS/MS prediction across public resources while keeping molecular encoders, metadata conditioning, predictor heads, and downstream retrieval under one protocol. FlexMS establishes a fair evaluation playground which significantly lowers the barrier for integrating new predictive tools. Rather than solely optimizing for average scores, FlexMS augments aggregate accuracy with difficulty-aware diagnostics, providing actionable guidance on model selection across different compute constraints, data scales, and downstream retrieval objectives. Ultimately, FlexMS provides the community with a reproducible standard to identify which algorithmic conclusions are stable and which operating points are most viable in practice.

04.
arXiv (CS.AI) 2026-06-17

First, do NOHARM: towards clinically safe large language models

arXiv:2512.01241v3 Announce Type: replace-cross Abstract: Large language models (LLMs) are routinely used by physicians and patients for medical advice, yet their clinical safety profiles remain poorly characterized. We present NOHARM (Numerous Options Harm Assessment for Risk in Medicine), a 1,100-task benchmark of primary care-to-specialist consultation cases to measure the frequency and severity of harm from LLM-generated medical recommendations. NOHARM covers 10 specialties, with 12,747 expert annotations for 4,249 clinical management options. Across 28 LLMs, recommendations carried the potential for severe harm in up to 22.6% of cases, with errors of omission accounting for more than 80% of severe errors. In a randomized trial of 101 generalist physicians, human benchmark performance significantly improved with AI assistance, yet physicians remained far from realizing the potential of AI tools, frequently ignoring essential advice surfaced by AI. Safety performance tracked general-intelligence and medical-knowledge benchmarks across the full range of models but decoupled at the frontier. Despite strong performance on existing evaluations, widely used AI models can produce medical advice with the potential for severe harm at non-trivial rates, highlighting the importance of explicit measurement of clinical safety.

05.
arXiv (CS.CV) 2026-06-16

DreamX-World 1.0: A General-Purpose Interactive World Model

DreamX-World 1.0 is a general-purpose interactive text/image-to-video world model for controllable long-horizon generation. It supports camera navigation, revisits to previously observed regions, and promptable events across photorealistic, game-style, and stylized domains. Our data engine combines camera-accurate Unreal Engine rendering, action-rich gameplay recordings, and real-world videos with recovered camera geometry. For camera control, we introduce E-PRoPE, a lightweight variant of projective positional encoding that retains PRoPE's projective camera geometry while applying camera-aware attention to spatially reduced tokens. We convert a bidirectional video generator into a few-step autoregressive world model using causal forcing, DMD-style distillation, and long-rollout training. Training on self-generated long-horizon contexts exposes the model to its own generated history and reduces the style and color drift that accumulates across autoregressive chunks. Memory-Conditioned Scene Persistence retrieves earlier views through camera-geometry-based retrieval, while residual recycling makes the conditioning path less sensitive to imperfect memory latents. Event Instruction Tuning adds composable event control, and reinforcement learning alignment recovers camera control and visual quality after distillation. With mixed-precision DiT execution, residual reuse, 75\%-pruned VAE decoding, and asynchronous pipeline parallelism, DreamX-World 1.0 reaches up to 16\,FPS on eight RTX\,5090 GPUs. On our 5-second basic evaluation, DreamX-World 1.0 achieves a camera-control score of 73.75 and an overall score of 84.76, outperforming HY-WorldPlay 1.5 and LingBot-World in overall score, which achieve 80.79 and 80.45, respectively.

06.
arXiv (CS.CV) 2026-06-16

Rotational Symmetry based Object Pose Estimation from Point Clouds in the Absence of Known 3D Models

Object pose estimation is crucial to many industrial applications, with one example being automated spray painting using a robot. However, confidentiality concerns often limit access to high-quality 3D models, posing a significant challenge for point-cloud-based pose estimation. In such scenarios, rotational symmetry, a readily accessible characteristic of many industrial objects, can provide valuable prior information to facilitate pose estimation.In this paper, we propose a method that leverages the rotational symmetry commonly found in industrial objects to address the challenge caused by the absence of 3D models. The object pose is jointly estimated with point cloud refinement through an iterative optimization process. This optimization relies on a rotational symmetry constraint loss. To construct this loss, each 3D point is rotated according to the currently estimated pose, and multiple correspondences are identified using nearest-neighbor search by exploiting the rotational symmetry property. These correspondences are then used to compute the rotational symmetry constraint loss, which iteratively refines both the pose and the point cloud.By explicitly incorporating rotational symmetry into the optimization process, the proposed method achieves robust pose estimation and generalizes well across diverse object types. The proposed method is evaluated on a dataset specifically created for point clouds without known 3D models, consisting of four categories of synthetic objects and one real wheel hub collected from a production line. Experimental results demonstrate that the proposed method achieves performance comparable to methods that rely on known 3D models.

07.
arXiv (CS.AI) 2026-06-19

Review of Machine Learning Models for Solar Energetic Particle Prediction

arXiv:2606.19539v1 Announce Type: cross Abstract: Solar energetic particle (SEP) events have attracted increasing attention due to their significant radiation hazards for aviation, spacecraft electronics, and human missions beyond Earth's magnetosphere. From a scientific perspective, SEP events are intriguing because they arise from a set of physical processes extending from the solar surface and corona through the heliosphere, offering insight into particle acceleration and transport mechanisms that are widely applicable across astrophysics. Therefore, advancing our ability to understand and predict SEP events is essential both for deepening our knowledge of such mechanisms and for safeguarding space technologies and exploration. Traditionally, researchers have modeled SEPs using physics-based simulations and empirical methods. More recently, machine learning (ML) has emerged as a new tool for understanding and predicting SEP events. The purpose of this manuscript is to review the currently available ML models for SEP prediction, identify the datasets used for training, compare their architectures, inputs, and outputs, and, based on these insights, outline good practices and recommendations for future research.

08.
arXiv (CS.CV) 2026-06-12

JointEdit3D: Feed-Forward 3D Scene Editing in a Unified Latent Space

Existing 3D scene editing methods typically rely on per-scene optimization over explicit 3D representations or cascaded edit-and-reconstruct pipelines, resulting in high test-time cost, limited 3D awareness, and structural inconsistencies. To couple appearance synthesis and geometry prediction during editing, we build on a unified RGB-geometry reconstruction-generation latent space and adapt it to feed-forward 3D scene editing. The resulting framework, JointEdit3D, performs asymmetric latent inpainting by observing only a single edited RGB reference latent and generating the remaining RGB views and edited geometry latent under source-scene anchoring. JointEdit3D introduces a dedicated SceneAnchor Branch to inject source-scene structure without forcing direct copying, and adopts edit/background-aware losses to balance edited-region fidelity with unedited-content preservation. To address the lack of paired resources for standardized 3D scene editing evaluation, we introduce SceneEdit3D-15K, a dataset with 15K paired editing samples and renderer-provided 3D annotations, together with SceneEdit3D-Bench, a curated 100-sample benchmark. Experiments show that JointEdit3D improves edited-region quality and 3D structural completeness over prior baselines while maintaining competitive background preservation.

09.
arXiv (CS.CV) 2026-06-11

UI2Code^N: UI-to-Code Generation as Interactive Visual Optimization

UI-to-code aims to translate UI screenshots into executable front-end code. Despite progress with vision-language models (VLMs), most existing methods formulate UI-to-code as a single-pass generation, which mismatches real-world UI development that is inherently iterative and feedback-driven. We reformulate UI-to-code as an interactive visual optimization problem, where code generation is embedded in a closed-loop process of execution, visual inspection, and iterative refinement driven by rendered visual feedback. To address the non-differentiability of visual objectives and the noise of absolute visual evaluators, we propose Relative Visual Policy Optimization (RVPO), a preference-based reinforcement learning method that optimizes relative visual rankings among rendered candidates under execution feedback. We instantiate this paradigm in UI2Code^N, an open-source 9B model trained via continual pre-training, supervised fine-tuning, and reinforcement learning. Experiments demonstrate state-of-the-art performance on UI drafting, UI polishing, and UI editing benchmarks, even outperforming larger models, with performance consistently improving through iterative visual optimization. Our code and models are available at https://github.com/zai-org/UI2Code_N.

10.
arXiv (CS.CL) 2026-06-12

Agents' Last Exam

Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional domains. We argue that this gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows. This paper introduces Agents' Last Exam (ALE), a benchmark designed to evaluate AI agents on long horizon, economically valuable, real world tasks with verifiable outcomes. Developed in collaboration with 250+ industry experts, ALE covers non-physical industries defined with reference to O*NET / SOC 2018 (the U.S. federal occupational taxonomy). It is organized around a task taxonomy with 55 sub fields grouped into 13 industry clusters covering 1K+ tasks. Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is below 1%. ALE is designed as a living benchmark: its task pool grows continuously as new workflows and industries are onboarded. More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP relevant impact.

11.
arXiv (CS.AI) 2026-06-19

Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models

arXiv:2510.21978v2 Announce Type: replace-cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning and has become a standard post-training paradigm for contemporary language and vision-language models. However, the RLVR recipe introduces a significant risk of capability regression, in which models forget foundational skills after prolonged training without employing regularization strategies. We empirically confirm this concern, observing that open-source reasoning models suffer performance degradation on core capabilities such as perception and faithfulness. While imposing regularization terms like KL divergence can help prevent deviation from the base model, these terms are computed on the current task and therefore do not guarantee preservation of broader knowledge. Meanwhile, commonly used experience replay across heterogeneous domains makes it nontrivial to decide how much training emphasis each objective should receive. To address this, we propose RECAP-a replay strategy with dynamic objective reweighting for general knowledge preservation. Our reweighting mechanism adapts online using short-horizon signals of convergence and instability, shifting the post-training focus away from saturated objectives and toward underperforming or volatile ones. Our method is end-to-end and readily applicable to existing RLVR pipelines without training additional models or heavy tuning. Extensive experiments on benchmarks using Qwen2.5-VL-3B and Qwen2.5-VL-7B demonstrate the effectiveness of our method, which not only preserves general capabilities but also improves reasoning by enabling more flexible trade-offs among in-task rewards.

12.
arXiv (CS.AI) 2026-06-17

LongWebBench: Evaluating Structural and Functional Webpage Generation in Long-Horizon Settings

arXiv:2606.17727v1 Announce Type: new Abstract: Recent vision-language models (VLMs) have shown promising progress in generating webpages from visual inputs, yet existing evaluations mainly focus on short, single-screen, and largely static webpages. We introduce LongWebBench, a benchmark for evaluating long-horizon webpage generation from both structural and functional perspectives. LongWebBench contains 490 real-world long webpages for structural fidelity evaluation and 507 goal-oriented interaction tasks over 129 webpages for functional evaluation. It employs two complementary protocols: a multi-dimensional VLM-based metric for assessing long-range structural coherence, and a DOM-augmented agent-based pipeline for end-to-end functional verification. We further examine the automatic evaluation protocols through human agreement analysis. Experiments with state-of-the-art open-source and proprietary VLMs under single-image and multi-image settings reveal that structural fidelity degrades as webpage length increases, while visually plausible generations often fail to support executable multi-step interactions. These results highlight the need to evaluate long webpage generation beyond visual similarity, with executable interaction as a core criterion. Our code and data are available at https://github.com/zheny2751-dotcom/LongWebBench.

13.
arXiv (CS.LG) 2026-06-15

Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression

arXiv:2602.08324v5 Announce Type: replace Abstract: Chain-of-Thought (CoT) reasoning successfully enhances the reasoning capabilities of Large Language Models (LLMs), yet it incurs substantial computational overhead for inference. Existing CoT compression methods often suffer from a critical loss of logical fidelity at high compression ratios, resulting in significant performance degradation. To achieve high-fidelity, fast reasoning, we propose a novel EXTreme-RAtio Chain-of-Thought Compression framework, termed Extra-CoT, which aggressively reduces the token budget while preserving answer accuracy. To generate reliable, high-fidelity supervision, we first train a dedicated semantically-preserved compressor on mathematical CoT data with fine-grained annotations. An LLM is then fine-tuned on these compressed pairs via a mixed-ratio supervised fine-tuning (SFT), teaching it to follow a spectrum of compression budgets and providing a stable initialization for reinforcement learning (RL). We further propose Constrained and Hierarchical Ratio Policy Optimization (CHRPO) to explicitly incentivize question-solving ability under lower budgets by a hierarchical reward. Experiments on three mathematical reasoning benchmarks show the superiority of Extra-CoT. For example, on MATH-500 using Qwen3-1.7B, Extra-CoT achieves over 73\% token reduction with an accuracy improvement of 0.6\%, significantly outperforming state-of-the-art (SOTA) methods. Our source codes have been released at https://github.com/Mwie1024/Extra-CoT.

14.
arXiv (CS.CV) 2026-06-19

Current World Models Lack a Persistent State Core

World models are increasingly regarded as a decisive step toward artificial general intelligence, yet modeling the physical world demands more than rendering convincing frames on demand: it requires an internal world state that keeps evolving over time, decoupled from observation, so that objects endure and events run to their conclusions whether or not a camera is watching, much as the moon holds to its orbit when no one is looking. This requirement is a blind spot of existing benchmarks, which reward surface properties such as fidelity, motion, and camera controllability while never asking whether a generated world keeps evolving once it is unobserved. We introduce WRBench, the first systematic diagnostic benchmark that treats camera motion as an intervention on observability and resolves evaluation into a human-calibrated chain that asks whether the camera executes the requested interaction, whether the scene stays continuous and identifiable while in view, and whether a returning target remains consistent with the event that was set in motion. Across 9{,}600 videos from 23 models spanning four control paradigms, one finding proves stubborn: current systems maintain the observed world as a tracking shot, resuming a returning target in the state at which it was abandoned rather than advancing the event while it went unseen. Because this failure recurs across control paradigms, model families, and increments of scale, robust world-state evolution does not follow from cleaner imagery, tighter control, richer geometric priors, or sheer parameter count We therefore argue that the stability of the physical state kernel and the consistency of worldlines under viewpoint intervention should become first-class objectives of world-model design, so that a world model captures how the world will unfold rather than how the next frame appears.

15.
arXiv (CS.CL) 2026-06-16

SpatialWorld: Benchmarking Interactive Spatial Reasoning of Multimodal Agents in Real-World Tasks

Spatial reasoning is a foundational capability for multimodal large language models (MLLMs) to perceive and operate within the physical world. However, existing benchmarks predominantly rely on passive evaluation (e.g., static VQA) or simulator-specific pipelines, failing to assess general interactive spatial understanding. We introduce SpatialWorld, a unified benchmark designed specifically for evaluating the interactive spatial understanding of multimodal agents in complex real-world tasks. Integrating eight heterogeneous simulation backends under a shared, simulator-agnostic protocol, SpatialWorld features 760 human-annotated tasks across diverse domains (e.g., household routines, travel, social collaboration). Agents must solve tasks under vision-only partial observability, actively gathering egocentric visual evidence and expressing decisions via a unified, text-based action interface native to MLLMs. For reliable evaluation, each task includes a human-validated initial state, a reference trajectory, and a terminal-state verifier. Evaluating 15 advanced agents reveals that robust spatial task solving remains challenging: the strongest model, GPT-5, achieves an average task success rate (TSR) of only 17.4%, while the leading open-source model, Qwen-3.5, reaches 14.1%. Further analysis exposes a clear mismatch between task success and execution efficiency, alongside substantial domain-specific performance variations. These bottlenecks in active exploration and long-horizon planning position SpatialWorld as a rigorous testbed for future spatial agents.

16.
arXiv (CS.CL) 2026-06-11

Evolving Agents in the Dark: Retrospective Harness Optimization via Self-Preference

AI agents rely on a harness of skills, tools, and workflows to solve complex problems. Continually improving this harness is essential for adapting to new tasks. However, existing optimization methods typically require ground-truth validation sets, yet such labeled data is difficult to acquire in practical deployment settings. To address this problem, we introduce Retrospective Harness Optimization (RHO), a self-supervised method that optimizes the agent harness using only past trajectories. Specifically, RHO selects a diverse coreset of challenging tasks from past trajectories and re-solves them in parallel. The agent analyzes these rollouts using self-validation and self-consistency, then generates candidate harness updates and selects the most effective one by its own pairwise self-preference. We evaluate RHO across three diverse domains, spanning software engineering, technical work, and knowledge work. Notably, a single optimization round improves the pass rate on SWE-Bench Pro from 59% to 78% without any external grading. Furthermore, our analysis demonstrates that RHO effectively targets prior failure modes. As a result, the optimized harness alters the agent's behavior patterns and sustains higher accuracy during long-horizon sessions.

17.
arXiv (quant-ph) 2026-06-17

Probing PbTe-Pb nanowire devices with radio-frequency reflectometry

arXiv:2606.04544v2 Announce Type: replace-cross Abstract: We report the implementation of radio-frequency (rf) reflectometry on selective-area-grown PbTe-Pb nanowire devices on a CdTe substrate. These nanowires are predicted to host Majorana zero modes. We demonstrate the compatibility of the rf technique, including both resistive and capacitive sensing, with these nanowires. The effect of dielectric loss from the CdTe substrate is quantitatively characterized. Furthermore, the feasibility of rf reflectometry is verified under finite magnetic fields where zero-energy modes can emerge. Our results establish the fast control of PbTe quantum devices, paving the way for their applications in topological quantum computation.

18.
arXiv (quant-ph) 2026-06-16

Efficient Implementation of a Single-Qutrit Gate Set via Coherent Control

arXiv:2507.06860v2 Announce Type: replace Abstract: Qutrits offer the potential for enhanced quantum computation by exploiting an enlarged Hilbert space. However, the synthesis of high-fidelity and fast qutrit gates, particularly for single qutrits, remains an ongoing challenge, as it involves overcoming intrinsic constraints in quantum platforms. Here, we develop a novel framework for the efficient implementation of a single-qutrit gate set via coherent control, leveraging SU(3) dynamics while obviating platform-specific constraints such as those arising from the selection rule. As a proof-of-principle demonstration, we realize 35-ns qutrit Hadamard and X gates using a superconducting transmon, achieving an average fidelity of 99.5\%, as verified by randomized benchmarking. We further demonstrate two paradigmatic quantum circuits, which can be naturally extended to scalable qudit algorithms for phase estimation and parity check. In addition, we propose an SU(3)-based decomposition strategy for an arbitrary single-qutrit gate and numerically demonstrate its substantial efficiency improvement over conventional SU(2)-based protocols. By addressing the challenge of efficiently implementing single-qutrit gates, our protocol paves the way for realizing high-performance qutrit processors in diverse quantum platforms.

19.
arXiv (CS.CV) 2026-06-12

GeoCFNet: Geometry-Aware Confidence Field Network for Robot-Assisted Endoscopic Submucosal Dissection

Advanced surgical robotics has made robot-assisted endoscopic submucosal dissection (ESD) a promising approach for the en-bloc resection of large lesions, with the potential to reduce recurrence and improve long-term outcomes. However, the technical complexity and risk of complications in ESD demand stable and precise visual guidance to maintain an accurate dissection corridor and a safe tissue margin. Dense confidence fields provide an effective representation for this purpose by describing both the preferred dissection region and its spatial transition to surrounding tissue. However, reliable confidence field estimation remains challenging in dynamic endoscopic scenes due to smoke, specular highlights, tissue deformation, weak texture, and the thin geometric structure of the target region. To address these challenges, we formulate dissection guidance as a geometry-aware confidence field estimation problem and propose GeoCFNet, a geometry-aware confidence field network built on a pretrained DINOv3 backbone. GeoCFNet integrates a Token-Differentiated Fusion module to aggregate class-token context with dense patch representations, a SegFormer decoder for confidence regression, and Geometry-Aware Spatial Regularization (GASR) to preserve spatial coherence and local geometric transitions. Experimental results show that GeoCFNet achieves RMSE 0.0480, PSNR 27.1995, SSIM 0.3397, and CC 0.2466, indicating accurate and geometrically stable confidence field estimation for robot-assisted ESD guidance.

20.
arXiv (CS.CL) 2026-06-16

Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning

We introduce Nemotron 3 Ultra, a 550 billion total and 55 billion active parameter Mixture-of-Experts Hybrid Mamba-Attention language model. We pre-trained Nemotron 3 Ultra on 20 trillion text tokens, then extended the context length to 1M tokens, and post-trained using Supervised Fine Tuning (SFT), Reinforcement Learning (RL), and Multi-teacher On-Policy Distillation (MOPD). Nemotron 3 Ultra is our most capable model yet, employing multiple key technologies - LatentMoE, Multi Token Prediction (MTP), NVFP4 pre-training, multi-environment RLVR, MOPD, and reasoning budget control. Nemotron 3 Ultra achieves up to ~6x higher inference throughput as compared to state-of-the-art publicly available LLMs while attaining on-par accuracy. The state-of-the-art accuracy, high inference throughput, and 1M token context length make Nemotron 3 Ultra ideal for long-running autonomous agentic tasks. We open-source the base, post-trained, and quantized checkpoints, along with the training data and recipe on HuggingFace.

21.
arXiv (CS.CL) 2026-06-19

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models – DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) – both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) Manifold-Constrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability. We pre-train both models on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline that unlocks and further enhances their capabilities. DeepSeek-V4-Pro-Max, the maximum reasoning effort mode of DeepSeek-V4-Pro, redefines the state-of-the-art for open models, outperforming its predecessors in core tasks. Meanwhile, DeepSeek-V4 series are highly efficient in long-context scenarios. In the one-million-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2. This enables us to routinely support one-million-token contexts, thereby making long-horizon tasks and further test-time scaling more feasible. The model checkpoints are available at https://huggingface.co/collections/deepseek-ai/deepseek-v4.

22.
arXiv (CS.AI) 2026-06-12

MLUBench: A Benchmark for Lifelong Unlearning Evaluation in MLLMs

arXiv:2606.12809v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) are trained on massive multimodal data, making data unlearning increasingly important as data owners may request the removal of specific content. In practice, these requests often arrive sequentially over time, giving rise to the challenging problem of MLLM Lifelong Unlearning. However, most existing benchmarks are limited in scale and scope, failing to capture the complexities of MLLM lifelong unlearning. To fill this gap, we introduce the MLUBench, a large-scale and comprehensive benchmark featuring 127 entities across 9 classes under lifelong unlearning requests. We perform extensive experiments using MLUBench and reveal that existing unlearning methods suffer from severe, cumulative degradation. More critically, we further identify the unique challenge of this problem: unlike in unimodal models, MLLM lifelong unlearning is constrained by the need to preserve multimodal alignment. Continually unlearning from one modality could degrade the entire model. To alleviate this challenge, we propose LUMoE, an effective method. Experiments demonstrate that LUMoE significantly mitigates the degradation problem faced by baselines. The source code and the MLUBench dataset are open-sourced in https://github.com/lihe-maxsize/Lifelong_Unlearning_main.

23.
arXiv (CS.AI) 2026-06-17

LoopCoder-v2: Only Loop Once for Efficient Test-Time Computation Scaling

arXiv:2606.18023v1 Announce Type: cross Abstract: Looped Transformers scale latent computation by repeatedly applying shared blocks, but sequential looping increases latency and KV-cache memory with the loop count. Parallel loop Transformers (PLT) alleviate this cost through cross-loop position offsets (CLP) and shared-KV gated sliding-window attention, making loop count a practical design choice. We therefore study PLT loop-count selection through a gain–cost view: an extra loop may refine representations, but CLP also introduces a positional mismatch at each loop boundary. We instantiate this study by training LoopCoder-v2, a family of 7B PLT coders with different loop counts, from scratch on 18T tokens, followed by matched instruction tuning and evaluation. Empirically, the two-loop variant delivers broad gains over the non-looped baseline across code generation, code reasoning, agentic software engineering, and tool-use benchmarks, improving SWE-bench Verified from 43.0 to 64.4 points and Multi-SWE from 14.0 to 31.0 points. In contrast, variants with three or more loops regress, revealing a strongly non-monotonic loop-count effect. Our diagnostics show that loop 2 provides the main productive refinement, while later loops yield diminishing, oscillatory updates and reduced representational diversity. Because the CLP-induced mismatch remains roughly fixed as refinement gains shrink, the offset cost increasingly dominates. This gain–cost trade-off explains PLT's saturation at two loops and provides diagnostics for loop-count selection.

24.
arXiv (CS.CV) 2026-06-16

GroupToM-Bench: Benchmarking Group Theory of Mind and Nonlinear Social Emergence in MLLMs

True general intelligence requires not only a model of the physical world but also a social world model: the capacity to infer how individual mental states interact and crystallize into group-level outcomes. Despite notable progress in individual-level Theory of Mind (ToM) reasoning, existing multimodal large language models fail at this broader task. Collective behavior emerges non-linearly from social tensions, conformity dynamics, and structural constraints, meaning it cannot be recovered by merely summing individual intentions. We present GroupToM-Bench, the first multimodal benchmark for group-level ToM, built around a causal chain spanning micro-level BDI states (belief, desire, intention), meso-level group tension and structural constraints, and macro-level outcome prediction and mechanistic attribution. To probe this full arc, we develop a seven-level cognitive audit framework. Experiments reveal a gap between current models and human baselines, highlighting a failure to process social structures and non-linear collective dynamics.

25.
arXiv (CS.CV) 2026-06-18

Cosmos 3: Omnimodal World Models for Physical AI

We introduce Cosmos 3, a family of omnimodal world models designed to jointly process and generate language, image, video, audio, and action sequences within a unified mixture-of-transformers architecture. By supporting highly flexible input-output configurations, Cosmos 3 seamlessly unifies critical modalities for Physical AI – effectively subsuming vision-language models, video generators, world simulators, and world-action models into a single framework. Our evaluation demonstrates that Cosmos 3 establishes a new state-of-the-art across a diverse suite of understanding and generation tasks, demonstrating omnimodal world models as scalable, general-purpose backbones for embodied agents. Our post-trained Cosmos 3 models were ranked as the best open-source Text-to-Image and Image-to-Video models by Artificial Analysis, and the best policy model by RoboArena at the time the technical report was written. To accelerate open research and deployment in Physical AI, we make our code, model checkpoints, curated synthetic datasets, and evaluation benchmark available under the Linux Foundation's OpenMDW-1.1 License at https://github.com/nvidia/cosmos and https://huggingface.co/collections/nvidia/cosmos3. The project website is available at https://research.nvidia.com/labs/cosmos-lab/cosmos3.