Instructions to use hyunseoki/mh-v1b-math-9b-paonly with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use hyunseoki/mh-v1b-math-9b-paonly with Transformers:

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

pipe = pipeline("text-generation", model="hyunseoki/mh-v1b-math-9b-paonly")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("hyunseoki/mh-v1b-math-9b-paonly")
model = AutoModelForCausalLM.from_pretrained("hyunseoki/mh-v1b-math-9b-paonly", 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 hyunseoki/mh-v1b-math-9b-paonly with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "hyunseoki/mh-v1b-math-9b-paonly"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "hyunseoki/mh-v1b-math-9b-paonly",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/hyunseoki/mh-v1b-math-9b-paonly

SGLang

How to use hyunseoki/mh-v1b-math-9b-paonly 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 "hyunseoki/mh-v1b-math-9b-paonly" \
    --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": "hyunseoki/mh-v1b-math-9b-paonly",
		"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 "hyunseoki/mh-v1b-math-9b-paonly" \
        --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": "hyunseoki/mh-v1b-math-9b-paonly",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use hyunseoki/mh-v1b-math-9b-paonly with Docker Model Runner:
```
docker model run hf.co/hyunseoki/mh-v1b-math-9b-paonly
```

mh-v1b-math-9b-paonly

Part of the Meta-Harness v1b study — co-evolution GRPO (verl) where a shared-weight policy plays two roles: TA (grader / solver, single-turn) and Proposer (harness/scaffold editor via multi-turn SEARCH/REPLACE). Question: does co-evolving the policy and its evaluation harness help, and through which lever (policy vs scaffold)?


Domain	Math
Architecture	Qwen3_5ForCausalLM — 32 layers, hidden 4096
Precision	bf16 (merged from FSDP via `verl.model_merger`)
Notes	Ablation cell: PA-only (harness evolution axis only).

Collection: https://huggingface.co/collections/hyunseoki/meta-harness-v1b-policy-x-harness-co-evolution-6a5c9877f74920e8922808bc

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("hyunseoki/mh-v1b-math-9b-paonly")
model = AutoModelForCausalLM.from_pretrained("hyunseoki/mh-v1b-math-9b-paonly", torch_dtype="bfloat16")

Downloads last month: 16

Safetensors

Model size

9B params

Tensor type

BF16

Collection including hyunseoki/mh-v1b-math-9b-paonly

Meta-Harness v1b: Policy x Harness Co-Evolution

Collection

Co-evolution GRPO (verl, Qwen3.5): shared policy = grader/solver + harness-editor. IMO/paper-review/math. bf16 checkpoints + evolved scaffolds. • 20 items • Updated 6 days ago