The inability of LLMs of ask for clarification was exactly the flaw we encountered when testing them on open-ended problems, stated somewhat ambiguously. This w

The inability of LLMs of ask for clarification was exactly the flaw we encountered when testing them on open-ended problems, stated somewhat ambiguously. This was in the context of paradoxical situations, tested on DeepSeek-R1 and Claude-3.7-Sonnet. Blog post about our experiments: https://pankajpansari.github.io/posts/paradoxes/ https://pankajpansari.github.io/posts/paradoxes/