2 papers
cs.LG2025
Lost in the Middle: An Emergent Property from Information Retrieval Demands in LLMs
Nikolaus Salvatore, Hao Wang, Qiong Zhang
The performance of Large Language Models (LLMs) often degrades when crucial information is in the middle of a long context, a "lost-in-the-middle" phenomenon that mirrors the prima…
cs.CV2024
Just a Few Glances: Open-Set Visual Perception with Image Prompt Paradigm
Jinrong Zhang, Penghui Wang, Chunxiao Liu +5
To break through the limitations of pre-training models on fixed categories, Open-Set Object Detection (OSOD) and Open-Set Segmentation (OSS) have attracted a surge of interest fro…