TRAMS: Training-free Memory Selection for Long-range Language Modeling
This addresses inefficiencies in long-range dependencies for language models, though it appears incremental as it builds on existing architectures like Transformer-XL.
The paper tackles the problem of ineffective memories in long-range language modeling by proposing TRAMS, a plug-and-play strategy that selects tokens based on a simple metric to improve attention scores, resulting in performance improvements on benchmarks like WikiText-103 and enwik8 without extra training or parameters.
The Transformer architecture is crucial for numerous AI models, but it still faces challenges in long-range language modeling. Though several specific transformer architectures have been designed to tackle issues of long-range dependencies, existing methods like Transformer-XL are plagued by a high percentage of ineffective memories. In this study, we present a plug-and-play strategy, known as TRAining-free Memory Selection (TRAMS), that selects tokens participating in attention calculation based on one simple metric. This strategy allows us to keep tokens that are likely to have a high attention score with the current queries and ignore the other ones. We have tested our approach on the word-level benchmark (WikiText-103) and the character-level benchmark (enwik8), and the results indicate an improvement without having additional training or adding additional parameters.