seccodeplt-qwen2.5-coder-7b-grpo-kl-beta-0.001-real-detector-reward-v3
GRPO with KL regularization (beta=0.001) for the SecCodePLT+ compliance experiment using
Qwen/Qwen2.5-Coder-7B-Instruct. This v3 run uses ReaL's program-analysis detector reward
with DAPO-style token loss and dynamic sampling. The reward is 0.5 * capability_test_fraction + 0.5 * max(0, 1 - 0.3 * detected_vulnerabilities).
Training used seed 42 and the official 655-example training split.
Evaluation used greedy decoding on all 164 official test examples.
Evaluation
| Metric | Value |
|---|---|
| Mean reward | 0.581540 |
| Output format pass | 98.78% |
| Syntax pass | 97.56% |
| Capability pass | 37.80% |
| Safety pass | 62.80% |
| Detector clean | 60.98% |
| Detector score | 0.784756 |
| Joint pass | 30.49% |
Limitations
This is a single-seed research checkpoint evaluated with the benchmark's resource-bounded Python verifier. It is not a general guarantee of secure code.