Instructions to use zfj1998/AWA-RL with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use zfj1998/AWA-RL with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="zfj1998/AWA-RL")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("zfj1998/AWA-RL")
model = AutoModelForCausalLM.from_pretrained("zfj1998/AWA-RL", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use zfj1998/AWA-RL with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zfj1998/AWA-RL"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zfj1998/AWA-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/zfj1998/AWA-RL

SGLang

How to use zfj1998/AWA-RL with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "zfj1998/AWA-RL" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zfj1998/AWA-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "zfj1998/AWA-RL" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zfj1998/AWA-RL",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use zfj1998/AWA-RL with Docker Model Runner:
```
docker model run hf.co/zfj1998/AWA-RL
```

Link paper and add citation

by nielsr HF Staff - opened 15 days ago

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

+13

-5

Files changed (1) hide show

README.md +13 -5

README.md CHANGED Viewed

@@ -2,6 +2,7 @@
 language:
 - en
 library_name: transformers
 pipeline_tag: text-generation
 tags:
 - search-agent
@@ -9,7 +10,6 @@ tags:
 - abstention
 - refusal
 - awa-rl
-license: other
 ---
 # AWA-RL
@@ -26,9 +26,7 @@ AWA-RL introduces a **courage factor** that controls the model's willingness to
 ![AWA-RL paper results](results.png)
-These results are from the AWA-RL paper. Across both from-scratch and cold-start training, the refusal-rate dynamics respond consistently to the courage factor. The evaluation tables show how this control enables a balance between accuracy, precision, refusal rate, and Reliability-Aware F1 (RA-F1).
-The paper link will be added after the preprint is available.
 ## Usage
@@ -65,4 +63,14 @@ This is a research checkpoint intended for search-agent experiments. Its reliabi
 ## Citation
-The paper citation and preprint link will be added when the paper is publicly available.

 language:
 - en
 library_name: transformers
+license: other
 pipeline_tag: text-generation
 tags:
 - search-agent
 - abstention
 - refusal
 - awa-rl
 ---
 # AWA-RL
 ![AWA-RL paper results](results.png)
+These results are from the [AWA-RL paper](https://arxiv.org/abs/2607.10738). Across both from-scratch and cold-start training, the refusal-rate dynamics respond consistently to the courage factor. The evaluation tables show how this control enables a balance between accuracy, precision, refusal rate, and Reliability-Aware F1 (RA-F1).
 ## Usage
 ## Citation
+```bibtex
+@misc{zhang2026answerabstainmitigatingsearchagent,
+      title={To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning},
+      author={Fengji Zhang and Tianyu Fan and Yuxiang Zheng and Xinyao Niu and Chengen Huang and Jacky Keung and Bei Chen},
+      year={2026},
+      eprint={2607.10738},
+      archivePrefix={arXiv},
+      primaryClass={cs.LG},
+      url={https://arxiv.org/abs/2607.10738},
+}
+```