作者位次 #9/12ACL 2026Findings· Findings of ACL
Mitigating Position Bias in Transformers via Layer-Specific Positional Embedding Scaling
Zhenghua Wang · Yiran Ding · Changze Lv · Yixin Wu · Tianlong Li · Zhibo Xu · Muling Wu · Tianyuan Shi · Shizheng Li · Qi Qian · Xuanjing Huang · Xiaoqing Zheng
摘要
Large Language Models (LLMs) still struggle with the “lost-in-the-middle” problem, where critical information located in the middle of long-context inputs is often underrepresented or lost. While existing methods attempt to address this by combining multi-scale rotary position embeddings (RoPE),...