The Fragility of AI: How Prompts, Formats, and Elicitation Instruments Dictate LLM Performance
Recent AI research reveals a striking truth: LLM performance, memory retrieval, and stated preferences depend heavily on prompt formatting and evaluation instruments rather than stable internal capabilities.
AI资讯AI ResearchLLM Evaluation
81 views
