DuoGen framework enhances interleaved multimodal generation by data curation and architecture design, improving text quality and image consistency.

Min Shi, Xiaohui Zeng, Jiannan Huang et al.

PLACID leverages pre-trained video diffusion models with synthetic trajectories for multi-object compositing, achieving high fidelity and identity preservation.

Gemma Canet Tarrés, Manel Baradad, Francesc Moreno-Noguer et al.

Proposes risk management-based improvements for AI safety benchmarks, analyzing 210 tools with probabilistic and validity enhancements.

Cheng Yu, Severin Engelmann, Ruoxuan Cao et al.

Proposes DC-CoT with RL training to identify and execute parallel subtasks, reducing longest path length by 35-40%.

Arvind Mahankali, Kaiyue Wen, Tengyu Ma

Proposes SIDP, integrating reward-guided self-imitation with diffusion models, achieving 2.5× faster inference and improved robustness in visual navigation.

Runhua Zhang, Junyi Hou, Changxu Cheng et al.

Understanding generalization via embedding dimension and distributional convergence using Wasserstein distance and Lipschitz constants.

Junjie Yu, Zhuoli Ouyang, Haotian Deng et al.

Proposes FaceDefense, combining directional attribute editing and diffusion loss to generate imperceptible adversarial faces, enhancing defense against diffusion-based face swapping.

Yilong Huang, Songze Li

FNF method detects LLM origin via functional network activity consistency, high sample efficiency, robust.

Yiheng Liu, Junhao Ning, Sichen Xia et al.

Proposes TABROUGE, a training-free, deterministic LCS-based reward, improving table reasoning accuracy by 26.7 points.

Tung Sum Thomas Kwok, Xinyu Wang, Hengzhi He et al.

Study finds self-evolving LLMs rely on raw experience but struggle to utilize condensed experience effectively.

Weixiang Zhao, Yingshuo Wang, Yichen Zhang et al.

Using Gemini Deep Think, the study evaluated 700 'Open' Erdős problems, combining AI natural language verification with expert review, revealing problem status and AI limitations.

Tony Feng, Trieu Trinh, Garrett Bingham et al.

The paper proves static black-box evaluation cannot guarantee post-update alignment; larger models hide more latent biases, activated by single gradient steps.

Yavuz Bakman, Duygu Nur Yaldiz, Eleni Triantafillou et al.

Proposes Vision-DeepResearch, enabling multi-turn, multi-entity, multi-scale visual-text search for deep reasoning, supporting dozens of reasoning steps and hundreds of engine interactions.

Wenxuan Huang, Yu Zeng, Qiuchen Wang et al.

Proposes SUSTAINSCORE to quantify instruction-induced task performance drops; experiments show performance declines of 15-35%.

Yunjia Qi, Hao Peng, Xintong Shi et al.

PLANING combines triangle-Gaussian hybrid representation for streaming 3D reconstruction, achieving 18.52% improvement in Chamfer-L2 and 1.31dB PSNR over SOTA.

Changjian Jiang, Kerui Ren, Xudong Li et al.

Leviathan replaces input embedding with a learned vectorization layer, improving language modeling perplexity by 9% at 1.2B parameters, especially aiding rare words.

Reza T. Batley, Sourav Saha

Mechanistic framework models catastrophic forgetting as feature encoding transformations, validated via geometric analysis and Vision Transformer experiments.

Sergi Masip, Gido M. van de Ven, Javier Ferrando et al.

Proposes STARS, a Riemannian optimization-based inference-time method to enhance diversity by maximizing activation space volume with orthogonal steering directions.

Dongxuan Zhu, Ly Tran Ho Khanh, Andy Yat-Ming Cheung et al.

Proposes a cognitive friction metric based on hallucination detection in multimodal generative models, revealing semiotic ambiguities in spatial environments.

Javier Argota Sánchez-Vaquerizo, Luis Borunda Monsivais

OVD: trajectory matching with verbal scores reduces memory, improves Web QA and math reasoning by up to 25.7%.

Jing Xiong, Hui Shen, Shansan Gong et al.