This public repository contains the Qwen2.5-32B-Instruct LoRA SFT baseline adapter trained for the TTC Dense Verifier project.
Intended use
The adapter is a supervised fine-tuning baseline for code and systems-programming answer generation. It is not the main TTC verifier; the main verifier is the Qwen2.5-7B reward adapter released separately.
Base model
Base model: Qwen/Qwen2.5-32B-Instruct
Artifact type: LoRA SFT adapter
Training run: P12SFTB_20260604_133514, checkpoint 300
See:
training_config.yaml
evaluation_standard.md
model_manifest.json
Limitations
This is an adapter, not a full 32B model. Load it with the Qwen2.5-32B-Instruct base model.
