Vision-language-action (VLA) models transfer semantic knowledge acquired through vision-language pretraining to robot control, achieving strong performance across various benchmarks. However, most VLAs directly map observations to demonstrated actions without explicit supervision of intermediate physical state changes,...
Qian-Yu Lu, Jia-Run Li, Ce Zhang et al.· 0 citations
TimeBlind is presented, a diagnostic benchmark for compositional spatio-temporal understanding that demonstrates that even frontier models rely heavily on static visual shortcuts rather than genuine temporal logic, positioning TimeBlind as a vital diagnostic tool for next-generation video understanding.
Baiqi Li, Kangyi Zhao, Ce Zhang et al.· arXiv.org· 11 citations· ⚡4
VideoTreeSearch (VTS) is proposed, a framework that casts grounded LVQA as iterative self-correcting search over an adaptive temporal tree, and trains an agent to navigate the tree through four discrete operations: zoom_in, zoom_out, shift, and answer.
Ce Zhang, Ziyang Wang, Yu-Lu Pan et al.· arXiv.org· 0 citations
EgoMemReason is introduced, a comprehensive benchmark for week-long egocentric video understanding through memory-driven reasoning that evaluates three complementary memory types: entity memory, tracking how object states evolve and change across days; event memory, recalling and ordering activities separated by hours...
