epfl-dlab / spp-t0-3b-base

SPP-T0 — Base (3B)

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

Trained with Synthetic Persona Pretraining (SPP) from token zero: first-person reflections are inserted into the roughly 10% of annotated documents that carry one, throughout the entire pretraining run.

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: epfl-dlab/spp-t0-3b-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.

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 = "epfl-dlab/spp-t0-3b-base"
tok = AutoTokenizer.from_pretrained(repo)          # identical at every revision
model = AutoModelForCausalLM.from_pretrained(
    repo, revision="step-25000", dtype=torch.bfloat16, device_map="auto"
)

main always holds the finished model (step 254,313). 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.

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.

Links

• Paper: to be released
• Collection: https://huggingface.co/collections/epfl-dlab/spp-synthetic-persona-pretraining

License: to be finalised.