#sharded simulation17/05/2025
Microsoft and Salesforce Reveal Major Performance Drop of LLMs in Real Multi-Turn Conversations
New research from Microsoft and Salesforce shows that large language models experience a 39% performance drop when handling real multi-turn conversations with incomplete instructions, highlighting a key challenge in conversational AI.