VISOR method enhances LVLM efficiency by sparsely selecting vision-language interactions, reducing inference cost.

Adrian Bulat, Alberto Baldrati, Ioannis Maniadis Metaxas et al.

AgentRVOS combines SAM3 and MLLM for zero-shot video object segmentation, achieving leading performance.

Woojeong Jin, Jaeho Lee, Heeseong Shin et al.

CSTS enhances cross-environment AI detection stability through entity-relational abstraction, addressing schema perturbation collapse.

Abdul Rahman

c-CRAB dataset evaluates code review agents' abilities; current agents solve only 40% of tasks.

Yuntong Zhang, Zhiyuan Pan, Imam Nur Bani Yusuf et al.

3DCity-LLM enhances 3D city-scale perception with a coarse-to-fine feature encoding strategy, leveraging a 1.2M-sample dataset.

Yiping Chen, Jinpeng Li, Wenyu Ke et al.

Study shows LLMs struggle with test generation under software evolution, with pass rates dropping to 66% under semantic changes.

Sabaat Haroon, Mohammad Taha Khan, Muhammad Ali Gulzar

Introduced D2TC method to successfully evade IDS in IoT networks, enhancing attack success rate.

Islam Debicha, Tayeb Kenaza, Ishak Charfi et al.

SortedRL accelerates RL training for LLMs through online length-aware scheduling, enhancing efficiency and performance.

Yiqi Zhang, Huiqiang Jiang, Xufang Luo et al.

Introduces Graph Energy Matching (GEM), surpassing discrete diffusion models in molecular graph generation.

Michal Balcerak, Suprosana Shit, Chinmay Prabhakar et al.

ABot-PhysWorld: 14B diffusion transformer for physics-aligned robotic manipulation video generation, trained on 3 million manipulation clips.

Yuzhi Chen, Ronghan Chen, Dongjie Huo et al.

Proposes LLM Olympiad evaluation to address transparency and trust issues in model evaluation.

Jan Christian Blaise Cruz, Alham Fikri Aji

Proposes SIDReasoner, enhancing SID–language alignment and using outcome-driven reinforcement to improve reasoning in generative recommendation, boosting accuracy and interpretability.

Yingzhi He, Yan Sun, Junfei Tan et al.

Proposes DAK-UCB, integrating diversity and fidelity for online generative model selection using kernel UCB.

Donya Jafari, Farzan Farnia

PCR enhances KV cache reuse via prefix trees, layer-wise overlap, and prefetching, reducing TTFT by up to 2.47× in RAG systems.

Wenfeng Wang, Xiaofeng Hou, Peng Tang et al.

This study reveals CPU resource scarcity causes up to 7.11× increase in first-token latency in multi-GPU LLM inference, emphasizing CPU's critical role.

Euijun Chung, Yuxiao Jia, Aaron Jezghani et al.

PRISM framework employs importance sampling and neural prediction to efficiently learn humanoid robot safe-stoppability boundaries, enabling proactive safety monitoring.

Yifan Sun, Yiyuan Pan, Shangtao Li et al.

TrajLoom combines Grid-Anchor Offset, VAE, and flow matching to predict 81-frame dense trajectories, improving stability and realism.

Zewei Zhang, Jia Jun Cheng Xian, Kaiwen Liu et al.

CAPITU uses literary texts to design 59 automatically verifiable instructions, evaluating LLMs' instruction-following in Portuguese with cultural and morphological constraints.

Giovana Kerche Bonás, Roseval Malaquias Junior, Marcos Piau et al.

VideoDetective enhances long video understanding by integrating extrinsic query and intrinsic relevance, boosting VideoMME-long accuracy by 7.5%.

Ruoliu Yang, Chu Wu, Caifeng Shan et al.

UNITE achieves unified tokenization and latent diffusion with an autoencoder, reaching FID 2.12 on ImageNet.

Shivam Duggal, Xingjian Bai, Zongze Wu et al.