turtleyacht
13 hours ago
Looks like there is a memory hierarchy [1] even in LLMs.
Model can accept 1M tokens doesn't mean it can reason across those 1M tokens
2 pointsposted 14 hours ago
by annjose13 hours ago
Looks like there is a memory hierarchy [1] even in LLMs.