LTX-2 employs an asymmetric dual-stream Transformer with 14B (video) and 5B (audio) parameters, integrating cross-modal attention for synchronized audiovisual generation.

Yoav HaCohen, Benny Brazowski, Nisan Chiprut et al.

MedDialogRubrics evaluates LLMs' multi-turn diagnostic capabilities using a multi-agent system with 5,200 patient cases.

Lecheng Gong, Weimin Fang, Ting Yang et al.

LOST-3DSG employs low-dimensional word embeddings for lightweight open-vocabulary 3D scene graphs, enabling efficient dynamic object tracking.

Sara Micol Ferraina, Michele Brienza, Francesco Argenziano et al.

VLM4VLA uses minimal parameter fine-tuning to convert general VLMs into robot control policies, outperforming complex architectures.

Jianke Zhang, Xiaoyu Chen, Qiuyue Wang et al.

DGA-Net enhances camouflaged object detection using depth prompting and graph-anchor guidance, outperforming existing methods.

Yuetong Li, Qing Zhang, Yilin Zhao et al.

Falcon-H1R uses a hybrid model for efficient reasoning, achieving performance on par with larger models at only 7B parameters.

Falcon LLM Team, Iheb Chaabane, Puneesh Khanna et al.

Proposed Deferred Commitment Decoding (DCD) improves diffusion language model accuracy by 1.73%.

Yingte Shu, Yuchuan Tian, Chao Xu et al.

Introduces Hidden State Poisoning Attacks (HiSPA) against Mamba models, using short triggers to irreversibly overwrite hidden states and impair long-context retrieval, with experimental validation on ROBENCH-25.

Alexandre Le Mercier, Chris Develder, Thomas Demeester

Proposed feature-dependent noise models tested in DMControl and Meta-world; PbRL without explicit denoising outperforms noise-robust methods.

Yuxuan Li, Harshith Reddy Kethireddy, Srijita Das

OpenRT integrates 37 attack strategies, testing 20 models with an average success rate of 49.14%.

Xin Wang, Yunhao Chen, Juncheng Li et al.

ReToK integrates redundant token padding and hierarchical semantic regularization, significantly enhancing autoregressive image generation for long sequences.

Zixuan Fu, Lanqing Guo, Chong Wang et al.

SWE-Lego employs lightweight supervised fine-tuning with data augmentation and test-time scaling to achieve state-of-the-art software issue resolution.

Chaofan Tao, Jierun Chen, Yuxin Jiang et al.

Proposes a homogenization framework based on critical theory, introducing xeno-reproduction to enhance diversity in LLMs for AI safety.

Ian Rios-Sialer

ScienceDB AI uses LLM to recommend scientific datasets, achieving 30% accuracy improvement.

Qingqing Long, Haotian Chen, Chenyang Zhao et al.

Using DMFT to analyze high-dimensional random matrix-driven dynamics, revealing learning and generalization mechanisms.

Blake Bordelon, Cengiz Pehlevan

Fusion-SSAT leverages feature fusion of self-supervised reconstruction and global classification, achieving state-of-the-art cross-dataset deepfake detection with AUC 0.9613.

Shukesh Reddy, Srijan Das, Abhijit Das

SpaceTimePilot model achieves generative rendering of dynamic scenes with independent control over camera viewpoint and motion sequence.

Zhening Huang, Hyeonho Jeong, Xuelin Chen et al.

Using OpenForesight dataset, retrieval, and reinforcement learning, trained OpenForecaster 8B achieves competitive future prediction accuracy.

Nikhil Chandak, Shashwat Goel, Ameya Prabhu et al.

mHC projects residual mappings onto the doubly stochastic matrix manifold, restoring identity properties and enhancing large-scale training stability.

Zhenda Xie, Yixuan Wei, Huanqi Cao et al.

Proposes Encyclo-K, a knowledge statement-based dynamic benchmark, achieving contamination resistance and multi-knowledge assessment, with top model accuracy at 62.07%.

Yiming Liang, Yizhi Li, Yantao Du et al.