datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
surogate-onboarding-greeting
Surogate Onboarding Greeting
A tiny SFT dataset used by the Surogate guided onboarding. Every example maps
a varied user message to the identical assistant reply:
Hello, welcome to the Surogate platform!
Fine-tuning a small model (e.g. Qwen3-0.6B, LoRA, ~50 steps) on this dataset
makes it answer any input with the greeting — an instantly visible training
effect for first-time users.
~1031 examples, single train split
messages column: JSON-encoded OpenAI-style chat messages… See the full description on the dataset page: https://huggingface.co/datasets/surogate/surogate-onboarding-greeting.surogate-onboarding-greeting-eval
Surogate Onboarding Greeting — Eval Set
Held-out eval set for the Surogate guided onboarding. 24 instruction inputs
(none appear in the training set) each expect the exact answer:
Hello, welcome to the Surogate platform!
Used as an exact-match custom benchmark: the greeting-tuned model scores ~100%,
the untuned base model ~0% — an instantly readable training effect.
Columns: instruction, answer · single test split · synthetic, no PII.
Part of the Surogate platform onboarding.… See the full description on the dataset page: https://huggingface.co/datasets/surogate/surogate-onboarding-greeting-eval.