TransformerFAM: Feedback attention is working memory
This addresses the bottleneck of processing infinitely long sequences in Transformers, which is crucial for enhancing Large Language Models, though it appears incremental as it builds on existing Transformer frameworks.
The paper tackles the problem of Transformers' quadratic attention complexity limiting their ability to process long inputs by proposing Feedback Attention Memory (FAM), a novel architecture that uses a feedback loop to enable working memory, resulting in significant performance improvements on long-context tasks across model sizes like 1B, 8B, and 24B.
While Transformers have revolutionized deep learning, their quadratic attention complexity hinders their ability to process infinitely long inputs. We propose Feedback Attention Memory (FAM), a novel Transformer architecture that leverages a feedback loop to enable the network to attend to its own latent representations. This design fosters the emergence of working memory within the Transformer, allowing it to process indefinitely long sequences. TransformerFAM requires no additional weights, enabling seamless integration with pre-trained models. Our experiments show that TransformerFAM significantly improves Transformer performance on long-context tasks across various model sizes (1B, 8B, and 24B). These results showcase the potential to empower Large Language Models (LLMs) to process sequences of unlimited length.