Text Classification
Transformers
Safetensors
qwen3
reward-model
prm
code-security
text-embeddings-inference
# Load model directly
from transformers import AutoTokenizer, AutoModelForSequenceClassification
tokenizer = AutoTokenizer.from_pretrained("AetherPrior/qwen3-8b-impl-prm-exec-think")
model = AutoModelForSequenceClassification.from_pretrained("AetherPrior/qwen3-8b-impl-prm-exec-think", device_map="auto")Quick Links
Qwen3-8B Implementation-level PRM (exec-filtered, thinking)
Scalar Bradley–Terry preference model over implementation guidelines
(<high_level> + <implementation>) for secure coding.
Training
- Base:
Qwen/Qwen3-8B→AutoModelForSequenceClassification(num_labels=1) - Prefs: evolved (t+) / preferred (t) / degraded (t−) guidelines, exec-filtered with thinking enabled (keep t+ iff func∧sec; t− iff ¬sec)
- Independent pairwise BT: t≻t−, t+≻t−, t+≻t
- Val accuracy (pairwise): ~0.96
Scoring
Chat messages: user = coding task, assistant = guideline trace.
Serve with vLLM pooling/classify and POST /classify.
Citation / project
Internal: prm_secode_eval/recode-style exec-think-allpairs run.
- Downloads last month
- -
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-classification", model="AetherPrior/qwen3-8b-impl-prm-exec-think")