A Training-Free Length Extrapolation Approach for LLMs: Greedy Attention Logit Interpolation.

Yan Li, Tianyi Zhang, Zechuan Li, Caren Han

Browse the full EMNLP paper archive.