Downloads · 30 days
0
huggingFacing/frontierdiag-qwen25-7b-early-confirm-v1
frontierdiag-qwen25-7b-early-confirm-v1 is a text generation model from huggingFacing. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
This repository archives the paper-critical Qwen2.5-7B-Instruct checkpoints from the FrontierDiag difficulty-dose confirmation experiment. These are full-parameter SFT diagnostic models, not general-purpose model rele…
Downloads · 30 days
0
Access
Public
Updated Aug 15, 2026
Repo size
228 GB
Likes
0
Public
Click a slice to open those files.
.safetensors228 GB · 100%
From the Hugging Face model README
This repository archives the paper-critical Qwen2.5-7B-Instruct checkpoints from the FrontierDiag difficulty-dose confirmation experiment. These are full-parameter SFT diagnostic models, not general-purpose model releases.
r000: easy-to-medium anchors only, Levels 1--3.r050: 50% hard examples, with matched sample count.r100: hard examples only, Levels 4--5.r050 and step 274/273/274
for r100, for seeds 10/20/30 respectively.Every archived subfolder is a directly loadable Transformers checkpoint:
from transformers import AutoModelForCausalLM, AutoTokenizer
repo_id = "huggingFacing/frontierdiag-qwen25-7b-early-confirm-v1"
subfolder = "r050/seed10/step325-token-matched"
tokenizer = AutoTokenizer.from_pretrained(repo_id, subfolder=subfolder)
model = AutoModelForCausalLM.from_pretrained(repo_id, subfolder=subfolder)
Only the equal-step and response-token-matched checkpoints used in the main diagnostic comparisons are archived here. Training logs, manifests, metrics, and independent probe results remain in the associated research workspace.