1 paper · 1 filter
Jingguang Li, Yebo Wu, Zuyi Guo +7
Long-context reasoning is essential for complex and long-horizon tasks, yet the performance of large language models (LLMs) degrades as context length increases. Recent approaches…