#1 · Paper · Feb 28, 2025
FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference
Xunhao Lai, Jianqiao Lu, Yao Luo, Yiyuan Ma, Xun Zhou · arxiv.org
Best Papers for Computer Version in February 2025
Track this topic after sign-in.
Query conditions: topic=computer-version, publish_at in 202502, and type=paper
Top Picks
#2 · Paper · Feb 26, 2025
Towards Semantic Equivalence of Tokenization in Multimodal LLM
Shengqiong Wu, Hao Fei, Xiangtai Li, Jiayi Ji, Hanwang Zhang, Tat-Seng Chua, Shuicheng Yan · arxiv.org
#3 · Paper · Feb 25, 2025
You Only Sample Once: Taming One-Step Text-to-Image Synthesis by Self-Cooperative Diffusion GANs
Yihong Luo, Xiaolong Chen, Xinghua Qu, Tianyang Hu, Jing Tang · arxiv.org
#4 · Paper · Feb 18, 2025
MagicArticulate: Make Your 3D Models Articulation-Ready
Chaoyue Song, Jianfeng Zhang, Xiu Li, Fan Yang, Yiwen Chen, Zhongcong Xu, Jun Hao Liew, Xiaoyang Guo, Fayao Liu, Jiashi Feng, Guosheng Lin · arxiv.org
#5 · Paper · Feb 6, 2025
Ultra-Sparse Memory Network
Zihao Huang, Qiyang Min, Hongzhi Huang, Defa Zhu, Yutao Zeng, Ran Guo, Xun Zhou · arxiv.org
#6 · Paper · Feb 5, 2025
Shot2Story: A New Benchmark for Comprehensive Understanding of Multi-shot Videos
Mingfei Han, Linjie Yang, Xiaojun Chang, Lina Yao, Heng Wang · arxiv.org
FAQ
How is this best Computer Version papers page ranked?
This page ranks Computer Version papers by topic match, content-type filter, checkout momentum, and freshness. The ranking is recalculated as new items and engagement signals arrive, so the top results stay practical for current workflows instead of remaining static or purely chronological.
What does the time scope in this URL mean?
The time suffix in this URL defines the publish-date window used for ranking. Year paths include items in that year, and YYYYMM paths include one calendar month. This makes comparisons cleaner when you want a focused snapshot rather than an all-time aggregate.
Is this list paid placement?
No. This ranking is editorial and signal-driven, not sponsored placement. Attendemia evaluates public metadata, source context, and usage momentum to rank candidates. Payment does not buy position, so readers can interpret the list as a curation surface rather than advertising inventory.