Dataset · allenai
SFT instructions with personas and constraints
allenai/tulu-3-sft-personas-instruction-following
29,980 synthetic examples for training precise adherence to instructions with constraints from IFEval.
What's inside
The dataset contains 29,980 synthetic examples for SFT. Constraints are taken from the IFEval taxonomy, and instruction diversity is enhanced through personas. Data created by the Allen Institute for AI; language is English.
What you can do with it
- Fine-tune LLMs using SFT to follow constraints
- Train LoRA for instruction following
- Evaluate model robustness to strict user constraints
- Tasks
- Text generation
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 14.1K
- Likes
- 71
Models trained on it
- bartowski/allura-org_Q3-30B-A3B-Designant-GGUFText generation 2.4K
- mradermacher/Q3-Nilcall-32B-v0.1-i1-GGUF 1.8K
- nkthebass/tinybrainbot-100m-v5.1-instructText generation 1.7K
- nkthebass/tinybrainbot-360m-a75m-v5.1-chatText generation 1.2K
- nkthebass/tinybrainbot-53m-a9m-hybrid-instructText generation 1K
- nkthebass/tinybrainbot-360m-a75m-v5.1-instructText generation 993
- mradermacher/tinybrainbot-360m-a75m-v5-instruct-GGUF 909
- mradermacher/Q3-Nilcall-8B-v0.3-i1-GGUF 861
How to load it
from datasets import load_dataset
ds = load_dataset("allenai/tulu-3-sft-personas-instruction-following")