Flash-BoN combines timestep truncation, layer skipping, and activation proxies to optimize inference, achieving +8% AUC under fixed budgets.

Ruchit Rawal, Reza Shirkavand, Sayak Paul et al.

dOPSD enhances reasoning in diffusion language models by self-distillation, excelling on Dream and LLaDA datasets.

Phuong Tuan Dat, Qi Li, Xinchao Wang

This study investigates cross-task transferability in unified multimodal models, showing fully shared transformer architectures (like Lumina-DiMOO) achieve up to 9% accuracy gains in counting tasks.

Jiwon Kang, Heeji Yoon, Jaewoo Jung et al.

Proposes LLM-as-a-Tutor framework using pairwise comparison and atomic constraints to enhance non-verifiable RL.

Yujin Kim, Namgyu Ho, Sangmin Hwang et al.

The study explores GUI agents' reliance on pixels vs. structure, introducing the Perception-Fusion Gap metric.

Guijia Zhang, Yuxun Chen, Yuheng Qi et al.

Proposed RZDG dataset and a tracker-based geo-localization pipeline significantly improve roadwork zone detection and global positioning accuracy.

Zhiran Yan, Yutong Xin, S Shyam Shenoi et al.

AdaptiveSplat uses texture-aware pruning and adaptive Gaussian prediction to improve sparse 3D scene reconstruction without fine-tuning.

Badrinath Singhal, Srihari K G, Sreehari Iyer et al.

Combining quantized target model and speculative decoding, Qwen3.5-4B achieves 6.978x speedup on NVIDIA A10G.

Jaeyeon Kim, Jewon Lee, Bo-Kyeong Kim

Agentic IoT transforms IoT into cognitive systems via autonomous AI agents, enhancing real-time reasoning and adaptive planning.

Rümeysa Hilal Sevinç, Bahaeddin Türkoğlu, İbrahim Kök

Proposes a contract-based distributed MPC enabling anytime robot joining/leaving with guaranteed collision avoidance.

Sabrina Bodmer, Danilo Saccani, Melanie N. Zeilinger et al.

MRAC method significantly improves depth estimation robustness without adding parameters.

Sohag Roy, Rajesh Misra, Swami Shastravidyananda et al.

CrossHallu evaluates internal signals' cross-lingual and cross-domain generalization in LLMs, showing most models transfer hallucination signals effectively.

Aisha Alansari, Malak Alkhorasani, Hamzah Luqman

MANCE leverages manifold constraints for nonlinear concept erasure, significantly reducing information leakage while preserving other concepts.

Matan Avitan, Yoav Goldberg, Yanai Elazar

WSA1 employs 3D world-spatial-action joint modeling with only 6K hours of demonstration data, achieving 93% success in RoboTwin2.0 and +20% over SOTA in real tasks.

Jiahao Jiang, Jianing Zhang, Zhenhan Yin et al.

AdaptiveSD framework ensures stability and adaptability for LLM inference in CPU-constrained environments through multi-policy orchestration.

Sadra Saremi

ObjRetarget integrates human videos, polyhedral hand models, and anthropomorphic constraints to achieve high-precision robot motion retargeting.

Yuanchuan Lai, Qing Gao, Ziyan Liang et al.

GLLS introduces a training-free dual-stream framework combining global logical verification and active local search, achieving 84% accuracy on MMAD-QA.

Runzhi Deng, Yundi Hu, Yiming Zhong et al.

Proposes a Gaussian Splatting-based scene optimization with normal-guided depth propagation, outperforming SOTA in sparse-view 3D surface reconstruction.

Liang Han, Bangcai Wei, Junsheng Zhou et al.

BRAID models multi-modal reasoning as a unified MDP, jointly optimizing text and image generation, boosting multi-turn reasoning performance.

Zican Hu, Xuyang Hu, Yiming Liu et al.

CoRE-VLA uses conditional routing of experts for scalable and robust vision-language-action modeling, excelling in multi-task and long-horizon tasks.

Haozhe Zhang, Sixian Li, Yifei Zhang et al.