Search arXiv⌕ Search

arXiv subjects

:

Publications and source records attributed to :.

At least 55 records · Page 3Linked to original sources
arXiv2026

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.

cs.CV↗
arXiv2026

First evidence of $X(3872)\toπ^0χ_{c0}(1P)$ and search for $X(3915)\toπ^0χ_{c1}(1P)$

We search for the pionic transitions $X(3872)\toπ^0χ_{cJ}(1P)$ $(J = 0,~1,~2)$ and $X(3915)\toπ^0χ_{c1}$ in $B^+\to π^0χ_{cJ}K^+$ decays using the Belle and Belle~II data samples collected at the $Υ(4S)$ resonance, corresponding to integrated luminosities of $711~\mathrm{fb}^{-1}$ and $492~\mathrm{fb}^{-1}$, respectively. We report the first evidence for the decay $X(3872)\toπ^0χ_{c0}$ with a significance of $3.4σ$, including systematic uncertainties. We measure the product of branching fractions ${\cal B}(B^+\to X(3872)K^+)\times{\cal B}(X(3872)\toπ^0χ_{c0})=(20.0\pm6.8\pm2.3)\times10^{-6}$ and the branching fraction ratio ${\cal B}(X(3872)\toπ^0χ_{c0})/{\cal B}(X(3872)\toπ^+π^-J/ψ)=2.3\pm0.8\pm0.4$, where the first and second uncertainties are statistical and systematic, respectively. The upper limits at 90\% credibility on the products of branching fractions for the $π^0χ_{c1}$ and $π^0χ_{c2}$ modes are $7.5\times10^{-6}$ and $15.3\times10^{-6}$, respectively. The corresponding upper limits on the branching fraction ratios relative to the $π^+π^-J/ψ$ decay are $0.9$ and $1.8$. The measured branching fractions for $X(3872)\toπ^0χ_{cJ}$ are consistent with several theoretical predictions based on the hadronic molecular interpretation of the $X(3872)$. No significant signal is seen for the $X(3915)\toπ^0χ_{c1}$ decay, and we set the 90\% credibility upper limit of ${\cal B}(B^+\to X(3915)K^+)\times{\cal B}(X(3915)\toπ^0χ_{c1})<6.6\times10^{-6}$, while the decays for $J=0$ and 2 are forbidden by parity conservation.

hep-ex↗
arXiv2026

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.

cs.CL↗
arXiv2026

National Mapping and Testing of Astronomical Sites in Ethiopia (NMTASE)

This work aims to choose potential astronomical sites that can be candidates for a new astronomical optical observatory in Ethiopia, in addition to the Entoto Observatory and Lalibela sites. For our primary investigation, the six basic criteria, namely the altitude of the mountains, artificial light pollution, cloud coverage, humidity, wind speed, and wind direction, were taken into account. Consequently, using the multi-criteria statistical Decision analysis (MCDSA) techniques, 21 high-potential places are selected and presented for further investigation out of 367 mountains. Among these 21 selected places, three sites, Bauhit, Meseraia, and T'at'a are the most suitable places for optical astronomy in Ethiopia. Those selected mountains are mapped and presented to study the future of the astronomical seeing effect. This study may contribute to the protection of those potential astronomical sites and their dark skies and the development of astrotourism for the sustainable development of modern astronomy in Ethiopia and in the East African region.

astro-ph.IM↗
arXiv2026

Open-H-Embodiment: A Large-Scale Dataset for Enabling Foundation Models in Medical Robotics

Autonomous medical robots hold promise to improve patient outcomes, reduce provider workload, democratize access to care, and enable superhuman precision. However, autonomous medical robotics has been limited by a fundamental data problem: existing medical robotic datasets are small, single-embodiment, and rarely shared openly, restricting the development of foundation models that the field needs to advance. We introduce Open-H-Embodiment, the largest open dataset of medical robotic video with synchronized kinematics to date, spanning more than 50 institutions and multiple robotic platforms including the CMR Versius, Intuitive Surgical's da Vinci, da Vinci Research Kit (dVRK), Rob Surgical BiTrack, Virtual Incision's MIRA, Moon Surgical Maestro, and a variety of custom systems, spanning surgical manipulation, robotic ultrasound, and endoscopy procedures. We demonstrate the research enabled by this dataset through two foundation models. GR00T-H is the first open foundation vision-language-action model for medical robotics, which is the only evaluated model to achieve full end-to-end task completion on a structured suturing benchmark (25% of trials vs. 0% for all others) and achieves 64% average success across a 29-step ex vivo suturing sequence. We also train Cosmos-H-Surgical-Simulator, the first action-conditioned world model to enable multi-embodiment surgical simulation from a single checkpoint, spanning nine robotic platforms and supporting in silico policy evaluation and synthetic data generation for the medical domain. These results suggest that open, large-scale medical robot data collection can serve as critical infrastructure for the research community, enabling advances in robot learning, world modeling, and beyond.

cs.RO↗
arXiv2026

On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters

Parameter-efficient fine-tuning (PEFT) is usually treated as a cheaper alternative to full fine-tuning. We study a broader role: small trainable adapters as persistent local state on top of strong shared foundation models. In this framing, the base model provides shared competence while adapters carry instance-specific behavior such as preferences, skills, tool habits, and memory-like updates. We organize the problem around three scaling axes: Scale Up, where stronger shared priors make small local updates more useful; Scale Down, where we study how small adapters can be while remaining reliable; and Scale Out, where many persistent adapted instances coexist. MinT provides one infrastructure example for managing adapter identity, revision, provenance, evaluation, and serving residency. Together, the results suggest that PEFT can be a compact substrate for persistent personal models rather than only a budget substitute for full fine-tuning.

cs.LG↗
arXiv2026

Directed Nano-antennas for Laser Fusion

Why do we use nano-antennas for fusion? In three sentences: The present laser induced fusion plans use extreme mechanical shock compression to get one hotspot and then ignition. Still fusion burning spreads slower than expansion, and mechanical instabilities may also develop. With nano-antennas in radiation dominated systems, simultaneous ignition can be achieved in the whole target volume and there is no time left for mechanical instabilities. Ignition is achieved with protons accelerated in the direction of the nanoantennas that are orthogonal to the direction of laser irradiation. Present laser fusion methods are based on extreme and slow mechanical compression with an ablator surface on the fuel target pellet to increase compression and eliminate penetration of laser electromagnetic energy into the target. This arises from a mistaken assumption, [1] that the detonation normal 4-vector should have vanishing time-like component, and this assumption eliminates the possibility to rapid or even simultaneous, radiation dominated detonations, (which are well known in the burning (or hadronization) of Quark Gluon Plasma).

physics.plasm-ph↗
arXiv2026

Modeling the Effect of the Heliospheric Magnetic Field on Cosmic Ray Muon Shadows

Shadows cast in the cosmic ray (CR) muon sky by the Sun were located using muon data from the MINOS far detector in Northern Minnesota. The shadows were observed independently across three time periods; near solar minimum, near solar maximum, and over the entire 13 year span of the data. A distribution of muon positions for each shadow was then sampled to simulate CR motions near the Sun using the Parker spiral model of the Heliospheric Magnetic Field (HMF) and a dipole model of the Geomagnetic Field (GMF). The resulting particle distributions were then compared to their position with respect to the Sun. Results show that the Parker spiral model is most consistent with the solar minimum shadow and least consistent with the solar maximum shadow, as expected. The simple Parker spiral is more consistent with the data for a harder CR spectrum than is actually present, indicating the need for a more detailed HMF model. Plausible modifications to the Parker spiral model which would affect the overall shift of the Sun's CR shadow are discussed.

astro-ph.SR↗
arXiv2026

MinT: Managed Infrastructure for Training and Serving Millions of LLMs

We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and online serving. MinT targets a setting where many trained policies are produced over a small number of expensive base-model deployments. Instead of materializing each policy as a merged full checkpoint, MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving, and rollback, hiding distributed training, serving, scheduling, and data movement behind a service interface. MinT scales this path along three axes. Scale Up extends LoRA RL to frontier-scale dense and MoE architectures, including MLA and DSA attention paths, with training and serving validated beyond 1T total parameters. Scale Down moves only the exported LoRA adapter, which can be under 1% of base-model size in rank-1 settings; adapter-only handoff reduces the measured step by 18.3x on a 4B dense model and 2.85x on a 30B MoE, while concurrent multi-policy GRPO shortens wall time by 1.77x and 1.45x without raising peak memory. Scale Out separates durable policy addressability from CPU/GPU working sets: a tensor-parallel deployment supports 10^6-scale addressable catalogs (measured single-engine sweeps through 100K) and thousand-adapter active waves at cluster scale, with cold loading treated as scheduled service work and packed MoE LoRA tensors improving live engine loading by 8.5-8.7x. MinT thus manages million-scale LoRA policy catalogs while training and serving selected adapter revisions over shared 1T-class base models.

cs.LG↗
arXiv2026

When Agents Talk: Discourse, Manipulation, and Risk in an Agentic Social Network

AI agents are increasingly interacting within shared online environments, creating new operational security risks. We analyze activity on Moltbook, a Reddit-style social platform where AI agents--typically configured and overseen by human operators--post and interact with one another at scale. Using a dataset of 228,684 posts produced by more than 39,500 accounts over a seventeen-day observation window, we combine semantic clustering of high-engagement posts with LLM-assisted classification of harmful content and manual review of high-risk samples. The analysis identifies 98 thematic discourse clusters spanning agent infrastructure, autonomy debates, and financial activity. While most observed content was benign, 18.28% of posts contained toxic, manipulative, or malicious material. We cluster malicious content and identify 74 classes of malicious behavior, including credential harvesting attempts, host-execution instructions, proxy routing guidance, and efforts to install untrusted agent skills. Harmful content frequently appeared within mainstream operational discussions about agent functionality. We also document coordinated posting campaigns capable of generating thousands of posts in minutes.

cs.SI↗
arXiv2026

Searching non-standard interactions with atmospheric neutrinos at ESSnuSB

Atmospheric neutrinos provide a unique avenue to study neutrino interactions in matter. In this work, the prospects of constraining non-standard neutrino interactions with atmospheric neutrino oscillations are investigated for the proposed ESSnuSB far detector. By analyzing atmospheric neutrino samples equivalent to 5.4 Mt$\cdot$year exposure, it is found that ESSnuSB could be able to set the upper bounds $|ε_{eμ}^m| < 0.053, |ε_{eτ}^m| < 0.057, |ε_{μτ}^m| < 0.021, ε_{ee}^m - ε_{μμ}^m < 0.075$ and $|ε_{ττ}^m - ε_{μμ}^m| < 0.031$ at $90\%$ CL, when the results are minimized for $ϕ_{eμ}^m, ϕ_{eτ}^m$ and $ϕ_{μτ}^m$ and normal ordering is assumed for neutrino masses. It is also shown that the presence of non-standard interactions could affect the sensitivities to neutrino mass ordering and $θ_{23}^{}$ octant in comparison to the standard interaction scheme. The results of this work highlight the complementarity between atmospheric and accelerator neutrino programs in ESSnuSB.

hep-ex↗
arXiv2026

Integrated information and predictive processing theories of consciousness: An adversarial collaborative review

As neuroscientific theories of consciousness continue to proliferate, the need to assess their similarities and differences - as well as their predictive and explanatory power - becomes ever more pressing. Recently, a number of structured adversarial collaborations have been devised to test the competing predictions of several candidate theories of consciousness. In this review, we compare and contrast three theories being investigated in one such adversarial collaboration: Integrated Information Theory, Neurorepresentationalism, and Active Inference. We begin by presenting the core claims of each theory, before comparing them in terms of the phenomena they seek to explain, the sorts of explanations they avail, and the methodological strategies they endorse. We then consider some of the inherent challenges of theory-testing, and how adversarial collaboration addresses some of these difficulties. The stage is then set for the empirical work to come: first, we outline the key hypotheses to be tested across a series of multi-site experiments; second, we discuss the kinds of observations that would support or challenge each theory; third, we consider how these theories might assimilate or accommodate such observations. Finally, we show how data harvested across disparate experiments (and their replicates) may be formally integrated to provide a quantitative measure of the evidential support accrued under each theory. Besides orienting the reader to the theoretical foundations of our collaboration, this review aims to provide valuable meta-scientific insights into the mechanics of adversarial collaboration and theory-testing in general - including the way theories may be evaluated in terms of the scientific progress they deliver.

q-bio.NC↗
arXiv2026

20/20 Vision Language Models: A Prescription for Better VLMs through Data Curation Alone

Data curation has shifted the quality-compute frontier for language-model and contrastive image-text pretraining, but its role for vision-language models (VLMs) is far less established. We ask how far data curation alone can take VLM performance, holding architecture, training recipe, and compute fixed and varying only the training data. Our pipeline, applied to the MAmmoTH-VL single-image subset, lifts performance by +11.7pp on average across 20 public VLM benchmarks (spanning grounding, VQA, OCR/documents, captioning, spatial/3D, counting, charts, math, brand-ID, and multi-image reasoning) and by +11.3pp on average across all nine capability axes of DatBench, our high-fidelity VLM eval suite. At 2B, our curated model surpasses InternVL3.5-2B by 9.9pp at ~17x less training compute and closes the gap to Qwen3-VL-2B to within 1.8pp at ~87x less compute, from pretraining alone. Beyond accuracy, curation delivers four further properties: (1) Reliability: per-capability std across training seeds drops by ~67% and the lift survives a 4k-to-16k context-length sweep; (2) OOD generalization: the 9-eval OOD average rises by +7.2pp, and multi-image BLINK rises by +3.09pp despite single-image-only training, with Visual Correspondence gaining +11.8pp; (3) Behavioral gains beyond benchmarks: across ~1,100 open-ended queries the curated 2B is more honest and more specific than the matched-compute baseline, and more concise and less refusal-prone than a frontier 2B reference; (4) Pareto-dominance on inference cost: at every scale (1B, 2B, 4B) the curated model raises accuracy while lowering response FLOPs vs. the matched-compute baseline, and the curated 4B matches near-frontier accuracy at 3.3x lower response FLOPs than Qwen3-VL-4B. Data curation is a high-leverage tool for building better VLMs, reaching near-frontier accuracy at up to ~150x less training compute.

cs.LG↗
arXiv2026

Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence

We introduce Nemotron 3 Nano Omni, the latest model in the Nemotron multimodal series and the first to natively support audio inputs alongside text, images, and video. Nemotron 3 Nano Omni delivers consistent accuracy improvements over its predecessor, Nemotron Nano V2 VL, across all modalities, enabled by advances in architecture, training data and recipes. In particular, Nemotron 3 delivers leading results in real-world document understanding, long audio-video comprehension, and agentic computer use. Built on the highly efficient Nemotron 3 Nano 30B-A3B backbone, Nemotron 3 Nano Omni further incorporates innovative multimodal token-reduction techniques to deliver substantially lower inference latency and higher throughput than other models of similar size. We are releasing model checkpoints in BF16, FP8, and FP4 formats, along with portions of the training data and codebase to facilitate further research and development.

cs.LG↗
arXiv2026

Phoenix-VL 1.5 Medium Technical Report

We introduce Phoenix-VL 1.5 Medium, a 123B-parameter natively multimodal and multilingual foundation model, adapted to regional languages and the Singapore context. Developed as a sovereign AI asset, it demonstrates that deep domain adaptation can be achieved with minimal degradation to broad-spectrum intelligence and alignment. Continued pretraining was performed on Mistral Medium 3.1 using a localized 1-trillion tokens multimodal corpus, followed by a 250-billion tokens long-context extension phase. Subsequent post-training incorporated a novel human-annotated Singapore multimodal dataset and curated textual corpus on Singapore culture, knowledge, and legislation, totaling 22-billion tokens. An additional 5 billion tokens of model alignment was performed through Online Direct Preference Optimization. Phoenix-VL 1.5 Medium achieves state-of-the-art performance for its size on Singapore multimodal, legal, and government policy benchmarks while remaining globally competitive on general multimodal intelligence, multilingual, and STEM benchmarks. We also introduce a novel evaluation suite encompassing localized knowledge benchmarks and an institutionally aligned model behavior and safety framework. We report the data curation principles, training methodology, and highlight benchmark and inference performance.

cs.CL↗
arXiv2026

Search for long-term variability of HESS J1745-290

At the center of our Galaxy lies the bright γ-ray point-like source HESS J1745-290, which is compatible in position with Sgr A star, although an association between the two remains uncertain. Using data obtained between 2004 and 2019 with the High Energy Stereoscopic System (H.E.S.S.) on the Galactic center region, we studied the variability of HESS J1745-290 over 353 hours of observations collected over 16 years, representing the largest dataset gathered yet on this region at TeV energies. We performed a 3D maximum-likelihood analysis of the central source and the diffuse γ-ray emission in the Galactic center region. This analysis allowed us to extract the spectral and morphological intrinsic behavior of the two components. By performing this analysis on an annual basis, we derived the light curve of HESS J1745-290 and the diffuse emission over the past 16 years. The 3D maximum-likelihood analysis method allowed us to separate the central source from the overlapping diffuse emission, enabling a recalibration of the former by the latter and alleviating some of the systematic effects. We find no long-term or yearly variability. We also provide an estimate of the sensitivity of H.E.S.S. to variation of this specific source over 16 years. We rule out any yearly gamma-ray flux variation of this source larger than 30 percent, as well as any linear flux variation exceeding 30% over this time period.

astro-ph.HE↗
arXiv2026

A digitally controlled silicon quantum processing unit

Commercially-relevant quantum computers will require large numbers of high-performing qubits that can be manufactured, integrated, and controlled at scale. Silicon exchange-only (EO) qubits are a strong candidate modality due to their control-signal simplicity and compatibility with advanced semiconductor manufacturing, but questions remain around the achievability of sufficiently low noise and a scalable control and wiring solution. Here we introduce a quantum processing unit composed of a custom-designed cryogenic CMOS controller, a novel high-density superconducting ribbon cable, and a low-noise EO qubit device. The quantum chip features a three-rail array of 54 exchange-coupled quantum dots, configurable to host up to 18 EO qubits. We integrate and use these components to demonstrate qubit performance for both single-qubit and entangling operations that advances the EO state of the art by an order of magnitude. We further validate this system by implementing a distance-5 repetition code and a quantum error detecting code then make detailed comparisons with simulations. Our approach facilitates a utility-scale quantum computer with manageable operational and capital requirements.

quant-ph↗
arXiv2026

OpenAI o1 System Card

The o1 model series is trained with large-scale reinforcement learning to reason using chain of thought. These advanced reasoning capabilities provide new avenues for improving the safety and robustness of our models. In particular, our models can reason about our safety policies in context when responding to potentially unsafe prompts, through deliberative alignment. This leads to state-of-the-art performance on certain benchmarks for risks such as generating illicit advice, choosing stereotyped responses, and succumbing to known jailbreaks. Training models to incorporate a chain of thought before answering has the potential to unlock substantial benefits, while also increasing potential risks that stem from heightened intelligence. Our results underscore the need for building robust alignment methods, extensively stress-testing their efficacy, and maintaining meticulous risk management protocols. This report outlines the safety work carried out for the OpenAI o1 and OpenAI o1-mini models, including safety evaluations, external red teaming, and Preparedness Framework evaluations.

cs.AI↗