PRIME: Evaluating Prompt Resolution Under Incompatible Instructions in LLMs

Tehreem Javed, Shumaim Fatimah, Masooma Bakhtiari +2

Large language models (LLMs) often encounter conflicting prompts, although current instruction following benchmarks assess those meta-instructions in isolation, limiting the insigh…