Picture for Zoë Prins

Zoë Prins

Is my model perplexed for the right reason? Contrasting LLMs' Benchmark Behavior with Token-Level Perplexity

Add code
Mar 31, 2026
Viaarxiv icon