New paper! 🎉 As I was wrapping up my PhD, two leading AI researchers asked me in an interview: With advanced LLMs and fine-tuned NLP models today, is Out-Of-Distribution (OOD) still a challenge? They believed it wasn't! 🤔 Our work challenges this common perception.
Are NLP models – fine-tuned or few-shot LLMs – robust to domain shifts?🧐
🚨Spoiler🚨: no, but the picture is more nuanced than it seems.
📣In our new preprint, we study this fundamental question, revealing insights that current research might overlook🕵️♂️
1/11
