CVE-OSS

CVE-OSS is a 20B-parameter CVE analyst derived from the openaccess-ai-collective/gpt-oss-20b base model. It specializes in producing structured vulnerability briefs covering background, affected components, exploitation flow, impact, and mitigation guidance.

Files

The repository contains the merged BF16 weights (model-00001-of-00009.safetensors ... model-00009-of-00009.safetensors), tokenizer artifacts, and the chat template. Download via huggingface-cli download Kushalkhemka/CVE-OSS or the code sample below.

Quick Start

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "Kushalkhemka/CVE-OSS"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [
    {"role": "system", "content": "You are a seasoned CVE analyst."},
    {"role": "user", "content": "Provide an in-depth brief on CVE-2021-3712."},
]
input_ids = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
output = model.generate(input_ids, max_new_tokens=600, temperature=0.2)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Last updated: 2025-11-28T11:07:17.076117Z