EmDashBench: the only LLM eval where a perfect score is 0. Counts em dashes (U+2014) in LLM responses to prompts engineered to tempt them.
- Python 100%
| results | ||
| benchmark.py | ||
| lyrics_ballad.txt | ||
| README.md | ||
| results | ||
| benchmark.py | ||
| lyrics_ballad.txt | ||
| README.md | ||