Author: Distillio · License: cc-by-4.0 · Format: jsonl
0 downloads · 0 stars · 0 mirrors · size unknown · 1 file
Teaches when NOT to call a tool: 15k SFT + 9k DPO preference pairs.
Tags: tool-calling · agentic · preference · dpo
Trains and evaluates the specific skill of knowing when NOT to call a tool, not just how — 15k SFT examples plus 9k DPO preference pairs (chosen vs. rejected tool-use decisions).
Direct download: https://huggingface.co/datasets/nvidia/When2Call/resolve/main/train/when2call_train_sft.jsonl
Or via the API: curl -L -o file "https://huggingface.co/datasets/nvidia/When2Call/resolve/main/train/when2call_train_sft.jsonl"
Added 2026-08-11. Platform: Distillio — the open source training data collective.