Why exams intended for humans might not be good benchmarks for LLMs like GPT-4

---

Training data contamination and other factors mean LLMs like GPT-4 succeeding on human exams might not be a good measure of their abilities.Read More