@HelloVyom 1) It's an information retrieval problem, where models struggle to retrieve specific info from a large bloated context
2) The "Lost-in-the-Middle" phenomenon where LLMs tend to prioritize info at the beginning and the end of a conversation, often overlooking details in the middle
Just got rate limited on copilot and can't even use models like gpt-4.1 anymore. Horrible models, unfair rate limits and the incoming usage based billing. ggs goodbye GitHub copilot🫡