Hermes Agent No‑COT is a curated, instruction‑style dataset derived from the lambda/hermes-agent-reasoning-traces (GLM‑5.1 split).It contains 7,055 multi‑turn tool‑calling conversations that have been cleaned and restructured for supervised fine‑tuning of small, efficient agentic models – especially those that use function‑calling within a chat environment.
✅ Why "No‑COT"?Every chain‑of‑thought (
…) block has… See the full description on the dataset page:
https://huggingface.co/datasets/LiteMind/hermes-agent-NoCOT.
