Show HN: WLM – A 70B model trained to decode "I'm fine" with 94.7% accuracygithub.com·1 pts·gwillen85·0