论文广场 - AcademicHub

01.

arXiv (quant-ph) 2026-06-15 DOI: arXiv:2606.14456

QCI Connect: A Modular Full-Stack Quantum Computing Platform

作者:

arXiv:2606.14456v1 Announce Type: new Abstract: In a world of various competing quantum computing architectures, hardware-agnostic, full-stack platforms are necessary to bring the full power of quantum computing hardware to domain experts via the cloud. QCI Connect and its Software Development Kit provide a reference architecture for a full-stack platform with a modular design and open-source interface definitions, built to facilitate a community-driven application ecosystem. Here, we present its overall design and features, central interfaces, and lessons learned, both for users of the platform and as a reference guide for future developments.

阅读与讨论 → 访问原文 →

02.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.14734

BRIDGE: Biological Evidence Refinement and Heterogeneous Dynamic Gating for Gene Regulatory Networks

作者:

Ziyang Dong ↗Shanwen Tan ↗Hengchuang Yin ↗Wei Liu ↗Yifan Wang ↗Siyu Yi ↗Jiancheng Lv ↗Wei Ju ↗

arXiv:2606.14734v1 Announce Type: cross Abstract: Motivation: Gene regulatory network inference from single-cell RNA sequencing (scRNA-seq) data is important for uncovering cell-state-specific transcriptional programs. However, scRNA-seq measurements are sparse and noisy, and experimentally validated TF-target interactions remain limited, making reliable inference challenging. Although graph neural networks have advanced GRN prediction, existing methods often rely on biologically unconstrained graph augmentation, such as random edge perturbation, and insufficiently control information transfer between genes and cells. These limitations may distort regulatory structures and weaken robustness under noisy and weakly supervised settings. Results: To address these issues, we propose an innovative framework named Biological Evidence Refinement and Heterogeneous Dynamic Gating for Gene Regulatory Networks (BRIDGE). BRIDGE extracts gene and cell representations from the expression matrix and its matrix dual, and performs contrastive learning in the gene space and cell space between self and neighbors across the co-expression-refined regulatory view and the original graph. It then applies heterogeneous gated encoding to adaptively regulate information transfer between genes and cells, enabling robust transcription factor-to-target gene prediction. Experiments on benchmark datasets spanning three network types and seven cell types show that BRIDGE achieves state-of-the-art AUROC and AUPRC in most settings. In particular, on Specific networks, BRIDGE improves average AUPRC by 5% over the second-best baseline, GCLink. In cross-cell-type few-shot transfer, BRIDGE consistently outperforms GCLink and GENELink across all six target cell types. A case study on hESC further supports the biological relevance of the predictions, with 9 of the top 10 and 46 of the top 100 novel TF-target interactions validated by ChIPBase.

阅读与讨论 → 访问原文 →

03.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15333

Replay What Matters: Off-Policy Replay for Efficient LLM Reinforcement Unlearning

作者:

Zirui Pang ↗Chenlong Zhang ↗Haosheng Tan ↗Zhuoran Jin ↗Jiaheng Wei ↗Zixin Zhong ↗

LLM unlearning has emerged as a cost-effective alternative to full retraining for removing hazardous knowledge from pretrained models while preserving general utility. Recent RL-based methods such as RULE reformulate unlearning as learning a refusal behavior, but their on-policy optimization repeatedly samples from the same forget and retain/boundary prompts throughout training. We identify a critical inefficiency in this process: easy cases quickly converge and provide little useful gradient signal, while hard cases near the forget/retain boundary continue to produce low-reward rollouts that are discarded after a single use. To address this issue, we propose ReRULE, an off-policy replay enhancement for reinforcement unlearning. ReRULE stores low-reward hard-case rollout groups in a replay buffer during early GRPO training and reuses them in later stages through importance-sampled off-policy updates, redirecting computation toward boundary cases that still require learning. Theoretically, we show that ReRULE yields a tighter hard-case convergence bound than pure on-policy RULE. Empirically, ReRULE improves MUSE-Books Retain Quality from 46.3 to 56.2 while adding only 5–11% training time across benchmarks. Its limited improvement on the simpler TOFU setting further supports the intended conditional behavior: replay is most beneficial when the hard/easy disparity is pronounced.

阅读与讨论 → 访问原文 →

04.

arXiv (CS.CV) 2026-06-17 DOI: arXiv:2603.03485

Phys4D: Fine-Grained Physics-Consistent 4D Modeling from Video Diffusion

作者:

Haoran Lu ↗Shang Wu ↗Songling Liu ↗Jianshu Zhang ↗Maojiang Su ↗Guo Ye ↗Chenwei Xu ↗Lie Lu ↗Pranav Maneriker ↗Fan Du ↗Manling Li ↗Zhaoran Wang ↗…

Recent video diffusion models have achieved impressive capabilities as large-scale generative world models. However, these models often struggle with fine-grained physical consistency, exhibiting physically implausible dynamics over time. In this work, we present Phys4D, a pipeline for learning physics-consistent 4D world representations from video diffusion models. Phys4D adopts a three-stage training paradigm that progressively lifts appearance-driven video diffusion models into physics-consistent 4D world representations. We first bootstrap robust geometry and motion representations through large-scale pseudo-supervised pretraining, establishing a foundation for 4D scene modeling. We then perform physics-grounded supervised fine-tuning using simulation-generated data, enforcing temporally consistent 4D dynamics. Finally, we apply simulation-grounded reinforcement learning to correct residual physical violations that are difficult to capture through explicit supervision. To evaluate fine-grained physical consistency beyond appearance-based metrics, we introduce a set of 4D world consistency evaluation that probe geometric coherence, motion stability, and long-horizon physical plausibility. Experimental results demonstrate that Phys4D substantially improves fine-grained spatiotemporal and physical consistency compared to appearance-driven baselines, while maintaining strong generative performance. Our project page is available at https://sensational-brioche-7657e7.netlify.app/

阅读与讨论 → 访问原文 →

05.

arXiv (CS.CV) 2026-06-16 DOI: arXiv:2605.25803

ATV-Net: Adaptive Triple-View Network with Dynamic Feature Fusion

作者:

Sheng-Wei Chan ↗Hsin-Jui Pan ↗Chun-Po Shen ↗Chia-Min Lin ↗Yung-Che Wang ↗Jen-Shiun Chiang ↗

Recent advances in semantic segmentation rely heavily on attention-based and transformer-style architectures that, while accurate, introduce considerable architectural complexity and computational cost. This paper asks whether a compact CNN-based segmentation head can remain competitive by adaptively selecting useful receptive-field evidence. We propose ATV-Net, an Adaptive Triple-View Network that attaches a lightweight head to a conventional backbone. The head organizes three complementary views – point-wise, neighborhood-level, and enlarged context – and fuses them through an Adaptive Decision Gate that generates image-dependent weights from global feature statistics. This allows the model to emphasize different receptive-field responses according to scene content, without dense attention or multi-scale aggregation. Experiments on Cityscapes and Pascal VOC 2012 show that ATV-Net achieves 80.31% mIoU on Cityscapes with ResNet-101 and 80.90% with ConvNeXt-Tiny, and 86.7% and 88.5% mIoU on Pascal VOC 2012, respectively, while requiring fewer GFLOPs than representative context-aggregation and attention-based heads. The results indicate that adaptive receptive-field selection remains a practical and effective design choice for CNN-based semantic segmentation.

阅读与讨论 → 访问原文 →

06.

arXiv (CS.CV) 2026-06-18 DOI: arXiv:2606.02800

Cosmos 3: Omnimodal World Models for Physical AI

作者:

NVIDIA ↗Aditi ↗Niket Agarwal ↗Arslan Ali ↗Jon Allen ↗Martin Antolini ↗Adeline Aubame ↗Alisson Azzolini ↗Junjie Bai ↗Maciej Bala ↗Yogesh Balaji ↗Josh Bapst ↗…

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.

阅读与讨论 → 访问原文 →

07.

arXiv (CS.AI) 2026-06-19 DOI: arXiv:2606.19635

Token Factory: Efficiently Integrating Diverse Signals into Large Recommendation Models

作者:

Xilun Chen ↗Shao-Chuan Wang ↗Baykal Cakici ↗Lukasz Heldt ↗Lichan Hong ↗Raghu Keshavan ↗Aniruddh Nath ↗Li Wei ↗Xinyang Xi ↗

arXiv:2606.19635v1 Announce Type: cross Abstract: Large Recommendation Models (LRMs) have demonstrated promising capabilities in industry-scale recommendation tasks. However, holistically integrating traditional signals into these transformer-based architectures effectively and efficiently remains a major challenge. Conventional approaches that "textualize" these signals directly or create discrete item representations often lead to excessively long prompts, substantial memory footprints, and high computational overhead. To overcome these limitations, we propose "Token Factory", a framework designed to transform traditional signals into "soft tokens" that can be directly processed by LRMs. This approach enables efficient integration and compression of heterogeneous input features, preventing prompt length explosion while enhancing model performance. We detail the architecture of Token Factory and present experimental results validating its effectiveness in a production-scale recommendation environment.

阅读与讨论 → 访问原文 →

08.

arXiv (CS.CL) 2026-06-11 DOI: arXiv:2606.12344

Claw-SWE-Bench: A Benchmark for Evaluating OpenClaw-style Agent Harnesses on Coding Tasks

作者:

Mengyu Zheng ↗Kai Han ↗Boxun Li ↗Haiyang Xu ↗Yuchuan Tian ↗Wei He ↗Hang Zhou ↗Jianyuan Guo ↗Hailin Hu ↗Lin Ma ↗Chao Xu ↗Guohao Dai ↗…

General-purpose agents such as OpenClaw are increasingly used as autonomous tool users, but their coding ability is difficult to measure under SWE-bench: a generic agent does not by itself satisfy the clean Docker workspace, patch, and prediction contract required for scoring. We introduce Claw-SWE-Bench, a multilingual SWE-bench-style benchmark and adapter protocol that makes heterogeneous agent harnesses, or claws, comparable under fair settings including a fixed prompt, runtime budget, workspace contract, patch extraction procedure, and evaluator. The full benchmark contains 350 GitHub issue-resolution instances across 8 languages and 43 repositories, drawn from SWE-bench-Multilingual and SWE-bench-Verified-Mini after future-commit cleanup. We also release Claw-SWE-Bench Lite for faster validation, which is an 80-instance subset selected by a cost-aware, rank-aware procedure over 17 calibration columns. On the full benchmark, OpenClaw with a minimal direct-diff adapter scores only $19.1\%$ Pass@1, whereas the full adapter reaches $73.4\%$ with the same GLM 5.1 backbone, showing that adapter design is essential for enabling OpenClaw-style harnesses to perform coding tasks effectively. Across an OpenClaw $\times$ nine-model sweep and a five-claw $\times$ two-model sweep, model choice changes Pass@1 by $29.4$ pp and harness choice by $27.4$ pp under fixed models; systems with similar accuracy can differ substantially in total API cost. Claw-SWE-Bench therefore treats harness and cost accounting as first-class axes of SWE-style coding-agent evaluation, providing both a full benchmark and a low-cost reference set for reproducible comparison. The data is available at https://github.com/opensquilla/claw-swe-bench and https://huggingface.co/datasets/TokenRhythm/Claw-SWE-Bench.

阅读与讨论 → 访问原文 →

09.

arXiv (CS.AI) 2026-06-12 DOI: arXiv:2606.12736

Benchmarking AI Agents for Addressing Scientific Challenges Across Scales

作者:

Tianyu Liu ↗Allen Xin Wang ↗Antonia Panescu ↗Lisa Xinyi Chen ↗Wenxin Long ↗Xinyu Wei ↗Yueqian Jing ↗Ziyao Zeng ↗Jihang Chen ↗Sihan Jiang ↗Ziqing Wang ↗Siyi Gu ↗…

arXiv:2606.12736v1 Announce Type: new Abstract: AI agents are increasingly being developed to accelerate scientific discovery, yet their practical capabilities in real research settings remain poorly understood. Existing benchmarks for AI agents rarely capture the complexity, heterogeneity, and extended reasoning required by scientific work, whereas benchmarks for scientific tasks often reduce research to static, direct problems and provide limited support for interactive evaluation. Here, we introduce SciAgentArena, a systematic benchmark for evaluating AI agents in real-world scientific research scenarios drawn from emerging needs across multiple domains. SciAgentArena comprises approximately 200 tasks with stepwise verification and an interactive, agent-agnostic environment for assessing diverse AI agents. Using this benchmark, we find that current agents can contribute effectively to well-specified data-analysis workflows, particularly when the task structure and evaluation criteria are clear. However, their performance remains uneven across scientific contexts: agents struggle to generate genuinely novel insights, sustain self-directed exploration, and formulate robust solutions for open-ended research questions. We further characterize common failure modes across agents and identify opportunities for improving their reliability, autonomy, and scientific reasoning. Together, SciAgentArena provides a practical framework for measuring progress in AI agents for science and for guiding the design of future agents capable of addressing complex scientific challenges. Full codes, tasks, and datasets can be accessed via this link: https://sciagentarena.github.io/.

阅读与讨论 → 访问原文 →

10.

arXiv (CS.AI) 2026-06-19 DOI: arXiv:2606.19613

StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns

作者:

Vlad Sobal ↗Shuo Yang ↗Yuting Zhang ↗Wei Xia ↗Stefano Soatto ↗

arXiv:2606.19613v1 Announce Type: cross Abstract: We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing. Unlike the prevailing fraction-of-tasks-solved metric, this matches real vibe-coding where sessions run dozens or hundreds of turns. In StaminaBench, agents implement a REST API server and modify it across a tunable number of procedurally generated follow-up change requests - 100 in our experiments, resulting in codebases of up to 6,000 lines. Tests are generated fully programmatically without LLM involvement, ensuring reproducibility and reliability; change sequences are drawn from either a hardcoded or LLM-driven sampler, both constrained to a structured action space to ensure changes are valid. The agent and the server run in an isolated environment and communicate with the benchmark through HTTP, making testing fully black-box and language-agnostic. We evaluate six agent harnesses paired with seven open-source LLMs across 20 scenarios of 100 turns each and find that: (1) all the tested models fail within 5-6 turns, confirming that vibe-coding-style programming without thorough testing produces bugs; (2) passing test feedback back to the agent and allowing it to retry improves passed turn count by up to 12x; and (3) a good harness is required for strong performance: stronger models exhibit up to a 6x gap between their best and worst harness, while weaker models fail with any harness. We release the benchmark and the generated tasks to enable further research into multi-turn coding agent behavior. Benchmark code and data: github.com/amazon-science/StaminaBench.

阅读与讨论 → 访问原文 →

11.

arXiv (CS.AI) 2026-06-19 DOI: arXiv:2606.19714

AURA: Adaptive Uncertainty-aware Refinement for LLM-as-a-Judge Auditing

作者:

Zilong Zhang ↗Yi-Ting Hung ↗Weiyi He ↗Junxi Zhang ↗Lei Ding ↗Chi-Kuang Yeh ↗

arXiv:2606.19714v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly used as judges for open-ended generation, as large-scale human evaluation is often expensive and difficult to scale, yet their preferences remain imperfect proxies for human judgment. Existing auditing pipelines often assume that a reliable subset of examples or clean supervision signals are available beforehand, for example from human annotation, heuristic filtering, or the outputs of strong judges. In LLM evaluation, this assumption is fragile: the initial split may inherit judge bias, while human verification is typically too scarce to define stable groups at scale. We propose AURA, an adaptive uncertainty–aware refinement framework for auditing pairwise LLM–as–a–judge decisions under selected human verification. AURA iteratively learns a human-consistency signal, propagates reliable evidence, and prioritizes uncertain comparisons for human review. The key idea is to treat trust in a judge as a latent quantity that is progressively refined as evidence accumulates. We provide a compact formulation, a stable refinement procedure, and a comprehensive evaluation on both synthetic and real pairwise LLM-answer data.

阅读与讨论 → 访问原文 →

12.

arXiv (CS.CV) 2026-06-16 DOI: arXiv:2606.17040

R2RDreamer: 3D-aware Data Augmentation for Spatially-generalized 2D Manipulation Policies

作者:

Xiuwei Xu ↗Haowen Sun ↗Angyuan Ma ↗Yiwei Zhang ↗Zhenyu Wu ↗Xiaofeng Wang ↗Bingyao Yu ↗Zheng Zhu ↗Jie Zhou ↗Jiwen Lu ↗

Spatial generalization is critical for imitation-learned manipulation policies, but achieving it typically requires scaling demonstrations across diverse object poses, robot configurations, and camera viewpoints. Data augmentation from a few source demonstrations offers a practical alternative to costly real-world collection. Simulation-based augmentation can create controllable variation, but requires complex environment and object setup and may introduce a sim-to-real gap. Recent real-to-real methods avoid these issues by jointly editing 3D observations and action trajectories from real demonstrations, yet they still rely on strong 3D scene parsing and geometry completion, and often produce observations tailored to 3D pointcloud policies rather than RGB-based 2D policies. We propose R2RDreamer, a real-to-real demonstration augmentation framework that preserves the geometric consistency of 3D action-observation editing while moving visual completion to 2D video space. Specifically, R2RDreamer first performs lightweight 3D augmentation by editing incomplete object pointclouds and end-effector trajectories in a shared 3D frame; it then projects the edited scene into masked image-space control videos with occlusion-aware reasoning and uses a dense-control image-to-video model to complete temporally coherent RGB observations. Experiments on spatially shifted manipulation tasks with both 2D diffusion-style policies and vision-language-action policies show that R2RDreamer improves spatial generalization from limited source demonstrations, with analyses validating the contributions of 3D editing, occlusion-aware projection, and video completion.

阅读与讨论 → 访问原文 →

13.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.15376

CoAgent: Concurrency Control for Multi-Agent Systems

作者:

Hongtao Lyu ↗Dingyan Zhang ↗Mingyu Wu ↗Xingda Wei ↗Haibo Chen ↗

arXiv:2606.15376v1 Announce Type: cross Abstract: Multi-agent LLM systems – coding agents, devops agents, document agents – now routinely run several agents in parallel against the same git tree, Kubernetes cluster, or document. As soon as two of them mutate shared state, they enter the regime classical concurrency control has studied for decades, but classical mechanisms fit LLM agents poorly. A single agent transaction spans minutes of inference, read sets are broad and opaque rather than statically inferable, and the live state agents act on admits neither fork nor buffer, so writes take effect the moment they execute. Locks block long inference intervals; OCC abort-and-retry discards minutes of work on every conflict. This paper builds concurrency control on a capability classical transactions lack: the LLM inside each agent can judge whether a conflicting write invalidates its plan, and can repair exactly the operations that depended on it. Control therefore turns advisory: the runtime informs, the agent repairs. Our protocol, MTPO (Monotonic Trajectory Pre-Order), fixes a serialization order at launch, serves each read the order-filtered value, and applies writes speculatively in place; a one-way notification asks an affected reader to re-judge and patch its plan, while the framework mechanically undoes and reorders misplaced writes through the saga-style inverse each tool registers in advance. At quiescence the run is serializable in the pre-decided order. We realize MTPO as CoAgent, toolcall middleware whose privileged ToolSmith grows footprint-declared, undoable tools online. On ten contended workloads, CoAgent stays within 5\% of serial correctness at a $1.4\times$ speedup and near-serial token cost, where 2PL and OCC surrender nearly all concurrency gains; on a bash-only target system, it grows a 25-tool library online and lifts the task pass rate from 45/71 to 63/71 at $0.80\times$ the time and $0.86\times$ the cost.

阅读与讨论 → 访问原文 →

14.

arXiv (CS.AI) 2026-06-19 DOI: arXiv:2606.19419

Playful Agentic Robot Learning

作者:

Junyi Zhang ↗Jiaxin Ge ↗Hanjun Yoo ↗Letian Fu ↗Zihan Yang ↗Yaowei Liu ↗Raj Saravanan ↗Shaofeng Yin ↗Justin Yu ↗Dantong Niu ↗Zirui Wang ↗Roei Herzig ↗…

arXiv:2606.19419v1 Announce Type: cross Abstract: Current agentic robot systems can write executable Code-as-Policy programs, observe feedback, and revise behavior across multiple attempts, but they remain largely task-driven: reusable skills are acquired only after explicit instructions. We study Playful Agentic Robot Learning, where an embodied coding agent uses self-directed play as a continual skill-learning stage before downstream tasks arrive. We introduce RATs, Robotics Agent Teams designed for play-time skill acquisition. During play, RATs proposes novel yet learnable exploratory tasks, plans and executes robot-code policies, verifies intermediate progress, diagnoses failures, retries with dense, step-level feedback, and distills successful executions into a persistent code skill library. At test time, the agent reuses relevant skills from this frozen library to help solve new tasks. Experiments in LIBERO-PRO and MolmoSpaces show that play-learned skills improve held-out downstream tasks over no-play and random-play baselines, with 20.6 and 17.0 percentage-point gains over CaP-Agent0 on LIBERO-PRO and MolmoSpaces, respectively. Moreover, the learned skills can be plugged into other inference-time Code-as-Policy agents by simply retrieving them into the context, improving RoboSuite and real-world transfer by 8.9 and 8.8 points, respectively, without finetuning the underlying model.

阅读与讨论 → 访问原文 →

15.

arXiv (CS.CV) 2026-06-15 DOI: arXiv:2606.14125

Conditioning Matters: Stabilizing Inversion and Attention in Diffusion Image Editing

作者:

Zheyuan Zhan ↗Hongchen Li ↗Can Wang ↗Yinfei Ma ↗Mingzhen Huang ↗Ruoshi Bai ↗Jiawei Chen ↗Siwei Lyu ↗Defang Chen ↗

Inversion-based image editing offers flexible and training-free control but still struggles with inversion accuracy and the trade-off between editing fidelity and background preservation. While recent methods improve inversion formulations or attention interactions, the role of textual conditioning in shaping diffusion dynamics and editing behavior remains underexplored. We show both empirically and theoretically that the precision of textual conditioning influences inversion stability by modulating the geometry of the diffusion velocity field, while also affecting the consistency of cross-branch attention during editing. These effects directly impact background preservation and semantic fidelity. Building on this analysis, we propose SimEdit, a conditioning-aware framework with two complementary components: (a) conditioning refinement, which constructs conditioning signals with improved semantic precision and structural alignment to facilitate stable inversion and consistent attention manipulation, and (b) token-wise cross-branch attention control, which separates edit-relevant and structure-preserving components and modulates them asymmetrically during attention manipulation. Extensive experiments on PIE-Bench demonstrate that SimEdit consistently improves both inversion reconstruction quality and editing performance over previous attention-manipulation approaches. Our code is available at https://github.com/zju-pi/SimEdit.

阅读与讨论 → 访问原文 →

16.

arXiv (CS.CV) 2026-06-11 DOI: arXiv:2509.11548

How Auxiliary Reasoning Unleashes GUI Grounding in VLMs

作者:

Weiming Li ↗Yan Shao ↗Jing Yang ↗Yujing Lu ↗Ling Zhong ↗Yuhan Wang ↗Min Yu ↗Tongxiao Ruan ↗Manni Duan ↗

Graphical user interface (GUI) grounding is a fundamental task for building GUI agents. However, general vision-language models (VLMs) struggle with this task due to a lack of specific optimization. We identify a key gap in this paper: while VLMs exhibit significant latent grounding potential, as demonstrated by their performance measured by Pointing Game, they underperform when tasked with outputting explicit coordinates. To address this discrepancy and bypass the high data and annotation costs of current fine-tuning approaches, we propose three zero-shot auxiliary reasoning methods. By providing explicit spatial cues such as axes, grids and labeled intersections as part of the input image, these methods enable VLMs to better articulate their implicit spatial understanding capabilities. We evaluate these methods on four GUI grounding benchmarks across seven open-source and proprietary VLMs. Experimental results show substantial gains from auxiliary reasoning. Mark-Grid Scaffold boosts Gemini-3.1-Pro from 11.72\% under direct inference to 95.20\% on ScreenSpot-v2, achieves state-of-the-art performance on ScreenSpot, and approaches the strongest fine-tuned methods on ScreenSpot-v2 and UI-I2E-Bench. Our code is available at https://github.com/liweim/AuxiliaryReasoning.

阅读与讨论 → 访问原文 →

17.

arXiv (CS.AI) 2026-06-16 DOI: arXiv:2606.15038

Fusion is not one-size-fits-all: Cross-Modal Representation Alignment for Time-to-Event Modeling

作者:

Zhemin Zhang ↗Weijie Chen ↗David Le ↗Amara Tariq ↗Alex Wallace ↗Matthew Stib ↗Juan Maria Farina ↗Chadi Ayoub ↗Reza Arsanjani ↗Imon Banerjee ↗

arXiv:2606.15038v1 Announce Type: new Abstract: Accurate time-to-event (TTE) prediction from multimodal clinical data remains challenging due to modality imbalance and distribution shift. We introduce a foundation model-driven framework for cross-modal representation alignment between CT imaging and longitudinal EHR data, designed to generalize across tasks and institutions. CT and EHR modalities are encoded independently using domain-specific foundation models and aligned in a shared latent space through four principled fusion strategies: late fusion, contrastive alignment, cross-attention, and co-attention. We evaluate two clinically distinct TTE tasks: pulmonary embolism (PE) mortality and cardiovascular disease (CVD) outcomes, on large-scale multi-institutional cohorts (PE: N=3,099 train; 1,098 internal; 435 external; CVD: N=2,951 train; 837 internal; 682 external). Fusion consistently improves concordance index by 1.5-5.4% over unimodal baselines when modalities contribute comparably. Overall, contrastive multimodal fusion, particularly with CLMBR representations, provided the most consistent and statistically robust improvements, especially for PE mortality prediction. For MACE, cross-attention (one-hot) achieved the highest internal performance and image-guided co-attention achieved the best external performance. We therefore introduce a generalizable foundation model-based cross-modal alignment framework and provide the first systematic analysis of fusion behavior under modality imbalance in TTE prediction. Our results establish task-aware multimodal alignment as a necessary design principle for robust generalization and scalable clinical deployment.

阅读与讨论 → 访问原文 →

18.

arXiv (CS.CL) 2026-06-19 DOI: arXiv:2605.16865

MixSD: Mixed Contextual Self-Distillation for Knowledge Injection

作者:

Jiarui Liu ↗Lechen Zhang ↗Yongjin Yang ↗Yinghui He ↗Yingheng Wang ↗Weihao Xuan ↗Zhijing Jin ↗Mona Diab ↗

Supervised fine-tuning (SFT) is widely used to inject new knowledge into language models, but it often degrades pretrained capabilities such as reasoning and general-domain performance. We argue this forgetting arises because fine-tuning targets from humans or external systems diverge from the model's autoregressive distribution, forcing the optimizer to imitate low-probability token sequences. To address this problem, we propose MixSD, a simple external-teacher-free method for distribution-aligned knowledge injection. Instead of training on fixed targets, MixSD constructs supervision dynamically by mixing tokens from two conditionals of the base model itself: an expert conditional that observes the injected fact in context, and a naive conditional that reflects the model's original prior. The resulting supervision sequences preserve the factual learning signal while remaining substantially closer to the base model's distribution. We evaluate MixSD on two synthetic corpora that we construct to study factual recall and arithmetic function acquisition in a controlled setting, together with established benchmarks for open-domain factual question answering and knowledge editing. Across multiple model scales and settings, MixSD consistently achieves a better memorization-retention trade-off compared to SFT and on-policy self distillation baselines, retaining up to 100% of the base model's held-out capability while maintaining near-perfect training accuracy, whereas standard SFT retains as little as 1%. We further show that MixSD produces substantially lower-NLL supervision targets under the base model and reduces harmful movement along Fisher-sensitive parameter directions. These results suggest that aligning supervision with the model's native generation distribution is a simple and effective principle for knowledge injection that mitigates catastrophic forgetting.

阅读与讨论 → 访问原文 →

19.

arXiv (CS.AI) 2026-06-17 DOI: arXiv:2606.17507

LLM-as-Judge in Education: A Curriculum-Grounded Marking Pipeline

作者:

Xiwei Xu ↗Chen Wang ↗Jacky Jiang ↗Phil Yang ↗Qian Fu ↗Mohan Dhall ↗Wenjie Zhang ↗Liming Zhu ↗

arXiv:2606.17507v1 Announce Type: new Abstract: Generative AI and large language models (LLMs) are increasingly applied to question generation and automated assessment. However, deploying LLMs in preparation for high-stakes exams requires more than prompt engineering; it demands software pipelines that systematically ground model outputs in authorised curriculum artefacts and marking guidelines issued by education authorities. This paper presents a curriculum-grounded, configurable LLM-as-Judge pipeline for question-level marking, co-developed with an industrial partner, to support exam preparation for university admission. The pipeline identifies the relevant topics, subtopics, and cognitive demand of a question, and assembles verifiable and authorised context to support LLM judgement. Curriculum intent is operationalised through concrete syllabus artefacts, including prescribed verbs and outcomes, performance band descriptors, glossary definitions, and marking-guideline principles. A staged LLM workflow is employed to first generate question-specific rubrics, capturing structured expectations of performance, and then derive and evaluate marking criteria used to allocate marks to student responses. This design improves consistency, transparency, and alignment with official marking practices. Preliminary evaluation shows that the proposed LLM-as-Judge pipeline delivers marking outcomes comparable to human tutors, while yielding justifications that are more traceable to authorised curriculum artefacts and marking standards. The pipeline has also been integrated into an online study platform, where early deployment data provide initial insights into operational usage and manual overrides.

阅读与讨论 → 访问原文 →

20.

arXiv (CS.CV) 2026-06-16 DOI: arXiv:2606.09150

Ultra Flash: Scaling Real-Time Streaming Video Generation to High Resolutions

作者:

Luxury ↗Jie Huang ↗Zihao Fan ↗Xiaoxiao Ma ↗Jun-hao Zhuang ↗Yuming Li ↗Zeyue Xue ↗Siming Fu ↗Haoran Li ↗Mingchen Zhong ↗Guohui Zhang ↗Shichen Ma ↗…

While recent autoregressive video diffusion models achieve remarkable streaming quality, they remain confined to low resolutions (e.g., 480P), leaving efficient, scalable, real-time high-resolution video generation a fundamental open challenge. To bridge this gap, we present Ultra Flash, a cascaded streaming framework capable of real-time high-resolution video generation. Ultra Flash achieves ~30 FPS at 1K resolution and ~18 FPS at 2K resolution on a single GPU through three key contributions: (1) an architecture-preserving T2V-to-TV2V super-resolution training paradigm coupled with an AIGC-oriented data degradation pipeline that effectively preserves the generative capability of the base model, enabling enhanced high-resolution detail when cascaded after mainstream low-resolution generative models; (2) a causal streaming latent upsampler paired with a high-resolution decoder, which enhances spatiotemporal coherence while enabling efficient latent spatial scaling and precise high-resolution decoding with negligible computational overhead; and (3) a cascade high-resolution streaming video generation optimization scheme that first performs hybrid-reward-enhanced sparse causalization and single-step distillation of the super-resolution model, then introduces cascaded streaming self-forcing preference optimization with dynamic cache management, jointly enhancing overall coherence, improving quality, and enabling real-time high-resolution streaming video generation. Extensive experiments demonstrate that Ultra Flash reliably produces ultra-high-resolution streaming video while maintaining state-of-the-art visual quality and superior efficiency. Project Page: https://xin1u.github.io/UltraFlash/

阅读与讨论 → 访问原文 →

21.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15893

BALTO: Balanced Token-Level Policy Optimization for Hallucination Mitigation

作者:

Ning Li ↗Zixuan Guo ↗Yan Xu ↗Wenbo Fei ↗Yifan Niu ↗Chang Luo ↗Yasheng Wang ↗Weiwen Liu ↗Yong Yu ↗Weinan Zhang ↗

Hallucinations remain a major obstacle to deploying large language models (LLMs) in knowledge-intensive settings, where generated responses must be faithfully grounded in provided evidence. Reinforcement learning (RL) is a promising direction for hallucination mitigation, but response-level faithfulness rewards suffer from a granularity mismatch: localized hallucinations can cause supported content to receive spurious penalties. Although recent work introduces fine-grained feedback such as claim-level verification and token-level rewards, unbalanced credit assignment can still induce length, verbosity, or optimization-noise biases. We propose BALTO, a Balanced Token-level Policy Optimization framework for hallucination mitigation. BALTO extracts checkable factual claims, verifies them against the reference context, and projects claim-level judgments to token-level labels. A balanced token-level credit assignment mechanism is introduced into the framework. This design redistributes probability mass from unsupported content toward faithful content, rather than suppressing the entire response. We systematically analyze the limitations of response-level rewards from a theoretical standpoint, and prove BALTO's advantages in training stability and optimization efficiency for hallucination mitigation. Experiments on ConFiQA, RAGTruth, and FinLLM-Eval show that BALTO achieves the highest faithfulness across all six model–benchmark settings and consistently outperforms existing post-training baselines in Q-Score, demonstrating a stronger faithfulness–informativeness trade-off.

阅读与讨论 → 访问原文 →

22.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15088

When the Same Musical Knowledge Forgets Differently: A Clean Probe of Pathway-Dependent Forgetting

作者:

Yu Liu ↗Zhiwei Yang ↗Wenxiao Zhang ↗Cong Cao ↗Fangfang Yuan ↗Kun Peng ↗Haimei Qin ↗Lei Jiang ↗Jin B. Hong ↗Hao Peng ↗Yanbing Liu ↗

A model can learn that the piano piece Für Elise is calm and reflective by listening to the audio or by reading a text description, but does it matter which route that knowledge took when it is later at risk of being forgotten? Forgetting research in multimodal models measures what knowledge is lost under adaptation, yet has not asked whether acquisition route affects how easily that knowledge is forgotten. We call this untested premise the Pathway-Invariant Assumption. Music understanding enables a clean test because a music clip and a canonical text description can be aligned to the same perceptual content, allowing the same knowledge unit to enter a model through listening or reading while the target remains fixed. Across multiple architecturally distinct audio-language models, we observe a consistent asymmetry: text-pathway knowledge is forgotten more than matched audio-pathway knowledge under identical adaptation pressure. To attribute this effect to route rather than confounds, we introduce the Paired Pathway Controlled Protocol (PPCP), a three-phase design that establishes matched pathway baselines, activates both pathways under symmetric supervision on the same knowledge pool, and applies identical forgetting pressure to both pathways. The gap is stable across models and gain-controlled analyses, persists when contradictory overwrite is replaced by correct-label cross-domain learning, remains under single-modality pressure, and is not removed by lightweight replay. Two independent routing-depth controls confirm that the effect is not explained by architectural depth, pointing to input representation as the dominant factor. Under PPCP, our results demonstrate that forgetting is highly route-dependent, establishing acquisition route as a new analytical dimension for forgetting research and multimodal system design.

阅读与讨论 → 访问原文 →

23.

arXiv (CS.CL) 2026-06-15 DOI: arXiv:2606.14243

Decoupled Mixture-of-Experts for Parametric Knowledge Injection

作者:

Baoqing Yue ↗Weihang Su ↗Qingyao Ai ↗Yichen Tang ↗Changyue Wang ↗Jiacheng Kang ↗Jingtao Zhan ↗Yiqun Liu ↗

Knowledge injection aims to equip large language models (LLMs) with external, domain-specific, or time-sensitive knowledge. Existing approaches typically face a trade-off between flexibility and integration: retrieval-augmented generation keeps knowledge outside the model but only provides prompt-level augmentation, whereas post-training based methods encode new knowledge into shared parameters but may introduce catastrophic forgetting, knowledge conflict, and costly updates. In this paper, we propose Decoupled Mixture-of-Experts (DMoE), a modular architecture for parametric knowledge injection that decouples both experts and the router from the base model. DMoE converts external knowledge corpora into independently updatable expert modules and uses a lightweight uncertainty-aware router to activate relevant experts only when the base model lacks sufficient knowledge during generation. To support efficient auto-regressive inference, DMoE attaches experts only to the final-layer feed-forward network, preserving KV-cache reuse while enabling parameter-level knowledge augmentation. Experiments on knowledge-intensive benchmarks show that DMoE consistently improves answer quality over retrieval and adapter-based baselines.

阅读与讨论 → 访问原文 →

24.

arXiv (CS.AI) 2026-06-15 DOI: arXiv:2512.02318

COGNITION: From Evaluation to Defense against Multimodal LLM CAPTCHA Solvers

作者:

Junyu Wang ↗Changjia Zhu ↗Yuanbo Zhou ↗Lingyao Li ↗Xu He ↗Mingkui Wei ↗Junjie Xiong ↗

arXiv:2512.02318v4 Announce Type: replace-cross Abstract: This paper studies how multimodal large language models (MLLMs) undermine the security guarantees of visual CAPTCHA. We identify the attack surface where an adversary can cheaply automate CAPTCHA solving using off-the-shelf models. We evaluate 7 representative MLLMs on 18 real-world CAPTCHA task types, measuring single-shot accuracy, success under limited retries, end-to-end latency, and per-solve cost. We further validate our findings through a supplemental external dataset and an adaptive-attacker setting with session memory, while also analyzing the impact of task-specific prompt engineering and few-shot demonstrations on solver effectiveness. We reveal that MLLMs can reliably solve recognition-oriented and low-interaction CAPTCHA tasks at human-like cost and latency, whereas tasks requiring fine-grained localization, multi-step spatial reasoning, or cross-frame consistency remain significantly harder for current models. By examining the reasoning traces of such MLLMs, we investigate the underlying mechanisms of why models succeed/fail on specific CAPTCHA puzzles and use these insights to derive defense-oriented guidelines for selecting and strengthening CAPTCHA tasks. To validate these principles, we present a proof-of-concept by hardening a vulnerable CAPTCHA type using our guidelines. We demonstrate that incorporating fine-grained localization and implicit counting reduces the success rate of state-of-the-art MLLMs from over 95\% to 0\%, confirming that structural changes can effectively mitigate the threat. We conclude by emphasizing the urgent need for CAPTCHA redesign as MLLM capabilities increasingly threaten existing defenses. Code Availability (https://doi.org/10.5281/zenodo.20406852).

阅读与讨论 → 访问原文 →

25.

arXiv (CS.CL) 2026-06-16 DOI: arXiv:2606.15007

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

作者:

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.

阅读与讨论 → 访问原文 →

探索全球前沿学术脉络