0

Tülu 3 SFT Mixture

Allen AI's flagship open SFT mixture combining new persona-driven prompts with curated public data for post-training a frontier-quality instruct model.

Open
Type
SFT Dataset
Publisher
Allen Institute for AI (Ai2)
Capabilities
SafetyCode GenerationMathInstruction Following
Runtime
hf_parquet
License
ODC-BY-1.0
Size
939k rows
Published
Jun 2023
Canonical
huggingface.co/datasets/allenai/tulu-3-sft-mixture

Cite

Notes

Only stored in your browser.

Lift evidence

4
EvalTools known to liftSource paper
Massive Multitask Language Understanding (MMLU)Tülu 3 SFT Mixture-
GSM8KTülu 3 SFT Mixture-
IFEvalTülu 3 SFT Mixture-
AlpacaEvalTülu 3 SFT Mixture-

Models

Notable models trained on it

Llama-3.1-Tulu-3-8BLlama-3.1-Tulu-3-70BLlama-3.1-Tulu-3-405B

Papers

2
introducesTulu 3: Pushing Frontiers in Open Language Model Post-TrainingTulu 3: Pushing Frontiers in Open Language Model Post-Training

Contributors

3
Hamish IvisonHannaneh HajishirziNathan Lambert