Not All Memories are Created Equal: Learning to Forget by Expiring
This addresses the computational bottleneck in sequence modeling for tasks requiring long-term memory, offering a scalable solution for reinforcement learning and language modeling.
The paper tackles the problem of efficiently scaling attention mechanisms to long sequences by proposing Expire-Span, a method that learns to retain important information and expire irrelevant memories, enabling Transformers to attend over tens of thousands of timesteps and achieving state-of-the-art performance on long-context tasks like character-level language modeling.
Attention mechanisms have shown promising results in sequence modeling tasks that require long-term memory. Recent work investigated mechanisms to reduce the computational cost of preserving and storing memories. However, not all content in the past is equally important to remember. We propose Expire-Span, a method that learns to retain the most important information and expire the irrelevant information. This forgetting of memories enables Transformers to scale to attend over tens of thousands of previous timesteps efficiently, as not all states from previous timesteps are preserved. We demonstrate that Expire-Span can help models identify and retain critical information and show it can achieve strong performance on reinforcement learning tasks specifically designed to challenge this functionality. Next, we show that Expire-Span can scale to memories that are tens of thousands in size, setting a new state of the art on incredibly long context tasks such as character-level language modeling and a frame-by-frame moving objects task. Finally, we analyze the efficiency of Expire-Span compared to existing approaches and demonstrate that it trains faster and uses less memory.