MBench evaluates long-term memory in video world models via entity, environment, and causal dimensions with 12 sub-metrics, using real long videos and VLM.

Shengjun Zhang, Zhang Zhang, Simin Huang et al.

Proposes Wan2.2 video diffusion model compression via few-step distillation and low-bit quantization, achieving superior efficiency.

Jinyang Du, Shenghao Jin, Ziqian Xu et al.

Introduces pause-and-think dataset and a 4B model fine-tuned with structured reasoning, achieving 58% accuracy with 59× fewer parameters than state-of-the-art models.

Shivam Singh, Saptarshi Majumder, Pratik Prabhanjan Brahma et al.

CARGO is a training-free routing framework using response agreement and Bayesian early stopping to decide when local models can answer reliably, supporting adjustable cloud offloading ratios.

Evan Chen, Shiqiang Wang, Kevin S Chan et al.

CAFOSat employs deep learning and GradCAM localization to refine annotations, covering 45,000 image patches across 20 US states for infrastructure-aware CAFO mapping.

Oishee Bintey Hoque, Nibir Chandra Mandal, Mandy L Wilson et al.

GNMR controls stability in low-precision language model training by comparing gradient norms with historical means.

Boao Kong, Weichen Jia, Engao Zhang et al.

Proposes semantic space matching using SSL features with Sinkhorn divergence, reducing ImageNet FID by 39×, improving one-step generative quality.

Hugues Van Assel, Edward De Brouwer, Saeed Saremi et al.

Proposes a kernel spectral method for semi-supervised regression with noisy proxy variables, achieving near-oracle rates under controlled noise conditions.

Kwangho Kim, Jisu Kim

EST-PRM framework stress-tests dense process reward models using three label-preserving transformations, revealing significant vulnerabilities across models.

Ibne Farabi Shihab, Fariya Afrin, Sanjeda Akter et al.

Proposed Grounded Decoding framework improves factual consistency in RAG without training.

Ibne Farabi Shihab, Fariya Afrin, Sanjeda Akter et al.

Zamba2-VL combines SSM and Transformer, achieving 10x faster inference at 1.2B-7B parameters.

Hassan Shapourian, Kasra Hejazi, Olabode M. Sule et al.

This work reveals that decoder attention capacity limits cause failure in deterministic reasoning; tool delegation helps surpass this barrier.

Dongxin Guo, Jikun Wu, Siu Ming Yiu

StressDream steers video world models by optimizing initial noise to enhance robust policy evaluation.

Junwon Seo, Sushant Veer, Ran Tian et al.

Lumos-Nexus employs a two-stage training and UPFB to bridge frequencies, boosting video fidelity and reasoning-driven generation.

Jiazheng Xing, Hangjie Yuan, Lingling Cai et al.

Introduces SOCO benchmark with over 1 million keypoint pairs across 100 categories, revealing that vision foundation models encode strong semantics but poorly transfer correspondences across categories.

Olaf Dünkel, Basavaraj Sunagad, Haoran Wang et al.

C4G introduces timestamp-conditioned learnable Gaussian tokens with transformer decoding, enabling efficient 4D scene reconstruction from monocular video without scene-specific optimization, reducing Gaussian count by orders of magnitude.

Mungyeom Kim, Minkyeong Jeon, Honggyu An et al.

LongTraceRL enhances long-context reasoning by generating complex multi-hop questions via knowledge graph random walks, using search trajectories for layered distractors, and entity-level rubric rewards, achieving significant improvements.

Nianyi Lin, Jiajie Zhang, Lei Hou et al.

Introduces context-dependent argumentation frameworks (CDAFs) with perspective activation, analyzing the complexity of strategic attack manipulation, establishing NP-completeness bounds.

Albert Sadowski, Jarosław A. Chudziak

Using Cultural Consensus Theory (CCT) to analyze LLM alignment in single- and multi-cultural settings, revealing errors in diversity representation and over-homogenization.

Krishna Pothugunta, John P. Lalor

This paper introduces the 'Disagreeing Rationales' framework, systematically analyzing how diverse human annotations and explanations impact hate speech detection, emphasizing the benefits of soft labels and rationales.

Benedetta Muscato, Beiduo Chen, Gizem Gezici et al.