← Previous · All Episodes · Next →
Fourier Position Embedding: Enhancing Attention's Periodic Extension for Length Generalization Episode 288

Fourier Position Embedding: Enhancing Attention's Periodic Extension for Length Generalization

· 21:34

|

🤗 Upvotes: 20 | cs.AI, cs.CL

Authors:
Ermo Hua, Che Jiang, Xingtai Lv, Kaiyan Zhang, Ning Ding, Youbang Sun, Biqing Qi, Yuchen Fan, Xue Kai Zhu, Bowen Zhou

Title:
Fourier Position Embedding: Enhancing Attention's Periodic Extension for Length Generalization

Arxiv:
http://arxiv.org/abs/2412.17739v1

Abstract:
Extending the context length of Language Models (LMs) by improving Rotary Position Embedding (RoPE) has become a trend. While existing works mainly address RoPE's limitations within attention mechanism, this paper provides an analysis across nearly all parts of LMs, uncovering their adverse effects on length generalization for RoPE-based attention. Using Discrete Signal Processing theory, we show that RoPE enables periodic attention by implicitly achieving Non-Uniform Discrete Fourier Transform. However, this periodicity is undermined by the spectral damage caused by: 1) linear layers and activation functions outside of attention; 2) insufficiently trained frequency components brought by time-domain truncation. Building on our observations, we propose Fourier Position Embedding (FoPE), which enhances attention's frequency-domain properties to improve both its periodic extension and length generalization. FoPE constructs Fourier Series and zero-outs the destructive frequency components, increasing model robustness against the spectrum damage. Experiments across various model scales show that, within varying context windows, FoPE can maintain a more stable perplexity and a more consistent accuracy in a needle-in-haystack task compared to RoPE and ALiBi. Several analyses and ablations bring further support to our method and theoretical modeling.


Subscribe

Listen to Daily Paper Cast using one of many popular podcasting apps or directories.

Apple Podcasts Spotify Overcast Pocket Casts
← Previous · All Episodes · Next →