README.md

---
library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen3-8B
tags:
- llama-factory
- full
- generated_from_trainer
datasets:
- rl-research/dr-tulu-sft-data
---

> [!NOTE]
> For full information, go check out the Dr Tulu paper [here](https://arxiv.org/abs/2511.19399).

<img src="https://huggingface.co/rl-research/DR-Tulu-SFT-8B/resolve/main/dr_tulu_logo.png" alt="Figure 1" width="500"/>


# DR Tulu SFT 8B

This is the SFT checkpoint of DR Tulu, an open deep research agent trained on top of [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B).

This model has undergone SFT training on [this dataset](https://huggingface.co/datasets/rl-research/dr-tulu-sft-data).
For more details on DR Tulu please **read our [paper](https://allenai.org/papers/drtulu)**!


# Inference and Usage

**This model has been trained for tool-use using the dr-agent-lib framework**.
As such, running it out of the box with HuggingFace or vLLM will not work well!

See [our github](https://github.com/rlresearch/dr-tulu) for more details on installation and how to run our model.
Or check out our [demo](https://dr-tulu.github.io/)!

# Evaluation Results

We provide evaluation instructions in [our github](https://github.com/rlresearch/dr-tulu).


| Benchmark | SQAv2 | HealthBench | ResearchQA | DeepResearch Bench | SimpleQA | 2Wiki | WebWalker | Average |
|:----------|:------:|:----------:|:---------:|:-------------------:|:------:|:-------:|-------:|-------:|
| [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) (naive rag) | 40.4 | 16.5 | 56.1 | 33.3 | 52.6 | 18.9 | 8.8 | 32.4 |
| [Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B) (our search pipeline) | 57.2 | 5.9 | 46.3 | 18.2 | 70.5 | 44.0 | 27.9 | 38.6 |
| [DR-Tulu-SFT-8B](https://huggingface.co/rl-research/DR-Tulu-SFT-8B) (**this model**) | 72.3 | 38.1 | 68.5 | 39.0 | 75.5 | 66.5 | 31.9 | 56.0 |
| [DR-Tulu-8B](https://huggingface.co/rl-research/DR-Tulu-8B) | **86.7** | **43.7** |  **71.1** | **41.8** | **80.1** | **68.0** | **39.1** | **61.5** |

For more baselines, explanations of this table, and analysis of results, check out the [Dr Tulu paper](https://allenai.org/papers/drtulu)!

# Intended uses & limitations

This model is licensed under Apache 2.0. It is intended for research and educational use in accordance with [Ai2's Responsible Use Guidelines](https://allenai.org/responsible-use).

## Training

The following hyperparameters were used during training:
- learning_rate: 4e-05
- train_batch_size: 1
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 16
- total_train_batch_size: 128
- total_eval_batch_size: 64
- optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 5

For futher details, check out the [Dr Tulu paper](https://allenai.org/papers/drtulu).

# Links
- 📝 [DR Tulu Paper](https://allenai.org/papers/drtulu)
- ⚙️ [DR Tulu demo](https://dr-tulu.github.io/)
- 💻 [DR Tulu code](https://github.com/rlresearch/DR-Tulu)
- 🤖 [DR Tulu collection](https://huggingface.co/collections/rl-research/dr-tulu)


# Citation
```
@article{shao2025dr,
  title={DR Tulu: Reinforcement Learning with Evolving Rubrics for Deep Research},
  author={Shao, Rulin and Asai, Akari and Shen, Shannon Zejiang and Ivison, Hamish and Kishore, Varsha and Zhuo, Jingming and Zhao, Xinran and Park, Molly and Finlayson, Samuel G and Sontag, David and others},
  journal={arXiv preprint arXiv:2511.19399},
  year={2025}
}
```