Instructions to use dlab-spp/t0-mt-3b-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dlab-spp/t0-mt-3b-base with Transformers:

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

pipe = pipeline("text-generation", model="dlab-spp/t0-mt-3b-base")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("dlab-spp/t0-mt-3b-base")
model = AutoModelForCausalLM.from_pretrained("dlab-spp/t0-mt-3b-base", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use dlab-spp/t0-mt-3b-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dlab-spp/t0-mt-3b-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dlab-spp/t0-mt-3b-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dlab-spp/t0-mt-3b-base

SGLang

How to use dlab-spp/t0-mt-3b-base 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 "dlab-spp/t0-mt-3b-base" \
    --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": "dlab-spp/t0-mt-3b-base",
		"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 "dlab-spp/t0-mt-3b-base" \
        --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": "dlab-spp/t0-mt-3b-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use dlab-spp/t0-mt-3b-base with Docker Model Runner:
```
docker model run hf.co/dlab-spp/t0-mt-3b-base
```

SPP-T0-MT — Base (3B)

Type: base (pretrained) model. Not instruction-tuned and ships no chat template.

Trained with SPP from token zero, followed by an additional reflection-focused midtraining stage resumed from the SPP-T0 checkpoint (replacing the final learning-rate cooldown).

Synthetic Persona Pretraining (SPP)

Synthetic Persona Pretraining (SPP) installs a target value persona during pretraining rather than only during alignment. Value-laden, first-person reflections, generated against a constitution, are appended to a subset of pretraining documents after a special <assistant> token. Attention masking and RoPE position aliasing keep the reflection from changing the continuation of the original document. This model is trained with SPP.

Instruction-tuned counterpart: dlab-spp/t0-mt-3b-instruct.

Model details

Architecture: Llama-3.2-3B-shaped, trained from scratch.
Tokenizer: the SmolLM2 tokenizer extended with an <assistant> marker and constitution tokens (vocabulary 49280).
Pretraining: ~500B tokens on a subset of the Olmo 3 Dolma 3 mixture, with SPP reflections inserted into the safety-annotated documents within it, followed by a reflection-focused midtraining stage on those annotated documents.

Training checkpoints

Intermediate checkpoints are published as git revisions on this repo, so any point in the trajectory can be loaded by passing revision=:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "dlab-spp/t0-mt-3b-base"
tok = AutoTokenizer.from_pretrained(repo)          # identical at every revision
model = AutoModelForCausalLM.from_pretrained(
    repo, revision="step-225000", dtype=torch.bfloat16, device_map="auto"
)

Revision	Stage	Tokens seen	Midtraining step	LR phase
`step-25000`	pretraining	~49.2B	—	stable
`step-50000`	pretraining	~98.3B	—	stable
`step-75000`	pretraining	~147B	—	stable
`step-100000`	pretraining	~197B	—	stable
`step-125000`	pretraining	~246B	—	stable
`step-150000`	pretraining	~295B	—	stable
`step-175000`	pretraining	~344B	—	stable
`step-200000`	pretraining	~393B	—	stable
`step-225000`	midtraining start	~442B	0 / 72,895	—
`step-250000`	midtraining	—	25,000 / 72,895	linear decay
`step-275000`	midtraining	—	50,000 / 72,895	linear decay
`step-297895`	midtraining	—	72,895 / 72,895	linear decay — same weights as `main`

main always holds the finished model (step 297,895). Only model weights are published — optimizer and RNG state are not included, so these revisions support evaluation, probing, and fine-tuning, but not exact resumption of the original run.

Revisions are named by global training step. Up to step-225000 this is the pretraining run of dlab-spp/t0-3b-base, and those revisions have the same weights as the revisions of the same name there. Midtraining started from dlab-spp/t0-3b-base step-225000 and ran 72,895 more steps, replacing the final learning-rate cooldown. The cooldown revisions of dlab-spp/t0-3b-base (step-240000, step-254313) are not part of this model's training history.

Intended use

Research on alignment and safety. As a base model it is meant for continuation, probing, or further fine-tuning; it is not instruction-tuned and can produce incorrect or unsafe content.

Citation

@misc{minder2026syntheticpersonapretrainingalignment,
      title={Synthetic Persona Pretraining: Alignment from Token Zero},
      author={Julian Minder and Viktor Moskvoretskii and Raghav Singhal and Difan Jiao and Andy Arditi and Shaobo Cui and Yiderigun Borjigin and Kartik Bali and Stefan Krsteski and Harsh Raj and Huu Nguyen and Jannik Brinkmann and Ashton Anderson and Roland Aydin and Robert West},
      year={2026},
      eprint={2608.13482},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2608.13482},
}

License: to be finalised.

Downloads last month: 164

Safetensors

Model size

3B params

Tensor type

BF16

Model tree for dlab-spp/t0-mt-3b-base

Adapters

1 model

Collection including dlab-spp/t0-mt-3b-base

Models - 3B

Collection

3B SPP models: bases publish their training trajectory as step-* revisions, plus persona-binding instruct counterparts. • 10 items • Updated Jul 31

Paper for dlab-spp/t0-mt-3b-base

Synthetic Persona Pretraining: Alignment from Token Zero

Paper • 2608.13482 • Published Aug 13 • 1

dlab-spp
/

t0-mt-3b-base