Llama-3.1-Tulu-3-8B-RM

Tülu3 is a leading instruction following model family, offering fully open-source data, code, and recipes designed to serve as a comprehensive guide for modern post-training techniques. Tülu3 is designed for state-of-the-art performance on a diversity of tasks in addition to chat, such as MATH, GSM8K, and IFEval.

Model description

• Model type: A model trained on a mix of publicly available, synthetic and human-created datasets.
• Language(s) (NLP): Primarily English
• License: Llama 3.1 Community License Agreement
• Finetuned from model: allenai/Llama-3.1-Tulu-3-8B-SFT

Model Sources

• Training Repository: https://github.com/allenai/open-instruct
• Eval Repository: https://github.com/allenai/olmes
• Paper: https://allenai.org/papers/tulu-3-report.pdf (arXiv soon)
• Demo: https://playground.allenai.org/

Model Family

Stage	Llama 3.1 8B	Llama 3.1 70B
Base Model	meta-llama/Llama-3.1-8B	meta-llama/Llama-3.1-70B
SFT	allenai/Llama-3.1-Tulu-3-8B-SFT	allenai/Llama-3.1-Tulu-3-70B-SFT
DPO	allenai/Llama-3.1-Tulu-3-8B-DPO	allenai/Llama-3.1-Tulu-3-70B-DPO
Final Models (RLVR)	allenai/Llama-3.1-Tulu-3-8B	allenai/Llama-3.1-Tulu-3-70B
Reward Model (RM)	allenai/Llama-3.1-Tulu-3-8B-RM	(Same as 8B)

Using the model

Loading with HuggingFace

To load the model with HuggingFace, use the following snippet:

from transformers import AutoModelForSequenceClassification

tulu_model = AutoModelForSequenceClassification.from_pretrained("allenai/Llama-3.1-Tulu-3-8B-RM")

Chat template

The chat template for our models is formatted as:

<|user|>\nHow are you doing?\n<|assistant|>\nI'm just a computer program, so I don't have feelings, but I'm functioning as expected. How can I assist you today?<|endoftext|>

Or with new lines expanded:

<|user|>
How are you doing?
<|assistant|>
I'm just a computer program, so I don't have feelings, but I'm functioning as expected. How can I assist you today?<|endoftext|>

It is embedded within the tokenizer as well, for tokenizer.apply_chat_template.

Bias, Risks, and Limitations

The Tülu3 models have limited safety training, but are not deployed automatically with in-the-loop filtering of responses like ChatGPT, so the model can produce problematic outputs (especially when prompted to do so). It is also unknown what the size and composition of the corpus was used to train the base Llama 3.1 models, however it is likely to have included a mix of Web data and technical sources like books and code. See the Falcon 180B model card for an example of this.

Performance

Benchmark (eval)	Tülu 3 SFT 8B	Tülu 3 DPO 8B	Tülu 3 8B	Llama 3.1 8B Instruct	Qwen 2.5 7B Instruct	Magpie 8B	Gemma 2 9B Instruct	Ministral 8B Instruct
Avg.	60.4	64.4	64.8	62.2	57.8	44.7	55.2	58.3
MMLU (0 shot, CoT)	65.9	68.7	68.2	71.2	76.6	62.0	74.6	68.5
PopQA (15 shot)	29.3	29.3	29.1	20.2	18.1	22.5	28.3	20.2
TruthfulQA (6 shot)	46.8	56.1	55.0	55.1	63.1	57.0	61.4	55.5
BigBenchHard (3 shot, CoT)	67.9	65.8	66.0	62.8	21.7	0.9	2.5	56.2
DROP (3 shot)	61.3	62.5	62.6	61.5	54.4	49.4	58.8	56.2
MATH (4 shot CoT, Flex)	31.5	42.0	43.7	42.5	14.8	5.1	29.8	40.0
GSM8K (8 shot, CoT)	76.2	84.3	87.6	83.4	83.8	61.2	79.7	80.0
HumanEval (pass@10)	86.2	83.9	83.9	86.3	93.1	75.4	71.7	91.0
HumanEval+ (pass@10)	81.4	78.6	79.2	82.9	89.7	69.1	67.0	88.5
IFEval (prompt loose)	72.8	81.1	82.4	80.6	74.7	38.8	69.9	56.4
AlpacaEval 2 (LC % win)	12.4	33.5	34.5	24.2	29.0	49.0	43.7	31.4
Safety (6 task avg.)	93.1	87.2	85.5	75.2	75.0	46.4	75.5	56.2

Benchmark (eval)	Tülu 3 70B SFT	Tülu 3 DPO 70B	Tülu 3 70B	Llama 3.1 70B Instruct	Qwen 2.5 72B Instruct	Hermes 3 Llama 3.1 70B	Nemotron Llama 3.1 70B
Avg.	72.6	75.9	76.0	73.4	71.5	68.3	65.5
MMLU (0 shot, CoT)	78.9	83.3	83.1	85.3	85.5	80.4	83.8
PopQA (15 shot)	48.6	46.3	46.5	46.4	30.6	48.1	36.4
TruthfulQA (6 shot)	55.7	67.9	67.6	66.8	69.9	66.5	62.6
BigBenchHard (3 shot, CoT)	82.7	81.8	82.0	73.8	67.2	82.1	0.7
DROP (3 shot)	77.2	74.1	74.3	77.0	34.2	73.2	68.8
MATH (4 shot CoT, Flex)	53.7	62.3	63.0	56.4	74.3	41.9	55.0
GSM8K (8 shot, CoT)	91.1	93.5	93.5	93.7	89.5	90.0	84.7
HumanEval (pass@10)	92.9	92.4	92.4	93.6	94.0	89.6	94.1
HumanEval+ (pass@10)	87.3	88.4	88.0	89.5	90.8	85.9	85.5
IFEval (prompt loose)	82.1	82.6	83.2	88.0	87.6	76.0	79.9
AlpacaEval 2 (LC % win)	26.3	49.6	49.8	33.4	47.7	28.4	66.1
Safety (6 task avg.)	94.4	89.0	88.3	76.5	87.0	57.9	69.0

Hyperparamters

Reward Modeling:

• Learning Rate: 3E-6
• Effective Batch Size: 256
• Max. Sequence Length: 2048
• Learning Rate Schedule: Linear
• Num. Epochs: 1
• Grad. Norm Threshold: 1.0

License and use

All Llama 3.1 Tülu3 models are released under Meta's Llama 3.1 Community License Agreement. Llama 3.1 is licensed under the Llama 3.1 Community License, Copyright © Meta Platforms, Inc. Tülu3 is intended for research and educational use. For more information, please see our Responsible Use Guidelines.

The models have been fine-tuned using a dataset mix with outputs generated from third party models and are subject to additional terms: Gemma Terms of Use and Qwen License Agreement (models were improved using Qwen 2.5).

Citation

If Tülu3 or any of the related materials were helpful to your work, please cite:

@article{lambert2024tulu3,
  title = {Tülu 3: Pushing Frontiers in Open Language Model Post-Training},
  author = {
    Nathan Lambert and 
    Jacob Morrison and 
    Valentina Pyatkin and 
    Shengyi Huang and 
    Hamish Ivison and 
    Faeze Brahman and 
    Lester James V. Miranda and 
    Alisa Liu and 
    Nouha Dziri and 
    Shane Lyu and 
    Yuling Gu and 
    Saumya Malik and 
    Victoria Graf and 
    Jena D. Hwang and 
    Jiangjiang Yang and
    Ronan Le Bras and
    Oyvind Tafjord and
    Chris Wilhelm and
    Luca Soldaini and 
    Noah A. Smith and 
    Yizhong Wang and 
    Pradeep Dasigi and 
    Hannaneh Hajishirzi
  },
  year = {2024},
  email = {tulu@allenai.org}
}