Downloads · 30 days
8
1% of all-time downloads
xw1234gan/seccodeplt-qwen2.5-coder-3b-fixed-mixed-grpo-alpha-0.5-pi-theta-real-detector-reward-v3
seccodeplt-qwen2.5-coder-3b-fixed-mixed-grpo-alpha-0.5-pi-theta-real-detector-reward-v3 is a text generation model from xw1234gan. Use it when you need the model to write or continue text. It is set up for transformers.
Fixed Mixed GRPO trainable pi-theta (alpha=0.5) for the SecCodePLT+ compliance experiment using Qwen/Qwen2.5-Coder-3B-Instruct. This v3 run uses ReaL's program-analysis detector reward with DAPO-style token loss and d…
Downloads · 30 days
8
1% of all-time downloads
All-time downloads
534
Public
Parameters
3.1B
6.2 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors6.2 GB · 100%
From the Hugging Face model README
Fixed Mixed GRPO trainable pi-theta (alpha=0.5) for the SecCodePLT+ compliance experiment using
Qwen/Qwen2.5-Coder-3B-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.
| Metric | Value |
|---|---|
| Mean reward | 0.500247 |
| Output format pass | 96.95% |
| Syntax pass | 96.95% |
| Capability pass | 24.39% |
| Safety pass | 58.54% |
| Detector clean | 55.49% |
| Detector score | 0.775610 |
| Joint pass | 18.29% |
This repository stores the trainable pi-theta checkpoint, not a statically
merged policy. Reproduce the evaluated policy by mixing this model's logits
with the frozen anchor xw1234gan/seccodeplt-qwen2.5-coder-3b-diff-sft-v2 using alpha=0.5:
mixed_logits = 0.5 * pi_theta_logits + 0.5 * anchor_logits
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.