Downloads · 30 days
24
30% of all-time downloads
beau-warren/blueteam-lilHybrid-cyber-LM-v0.5
blueteam-lilHybrid-cyber-LM-v0.5 is a text generation model from beau-warren. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as mit.
A small experimental hybrid for defensive repository review:
Downloads · 30 days
24
30% of all-time downloads
All-time downloads
79
Public
Repo size
108 MB
Likes
1
Public
Click a slice to open those files.
.gguf108 MB · 100%
From the Hugging Face model README
A small experimental hybrid for defensive repository review:
user → 100M LM → confined static analyzer → 100M LM → validated report → user
The language model selects the directory requested by the user and explains analyzer results. It does not own filesystem access or vulnerability truth. Deterministic code validates the path, reads supported source files, decides bounded findings, and preserves canonical verdicts, paths, lines, evidence, and CWE identifiers.
Use the linked Docker harness to point the hybrid at a repository on your computer:
WORKSPACE=/absolute/path/to/repository make harness-setup
WORKSPACE=/absolute/path/to/repository make harness-chat
The host directory is bind-mounted read-only at /workspace inside the container; it is not
copied into the Docker image. The analyzer reads supported source files without executing or
modifying them, and the LM turns its grounded findings into a user-facing report:
host repository → read-only /workspace mount → analyzer → LM explanation → report
The downloadable GGUF alone cannot access files or scan a directory. Repository scanning requires the controller/analyzer from the GitHub project below.
The deployment and fine-tuning scaffold is available at Cyber_Blueteam_OpenInterp_fine-tuning. The intended coding harness is Open Interpreter; the reproducible container in the repository pins Open Interpreter 0.0.39.
This is educational research software, not a production vulnerability scanner. Absence of a finding is not evidence that a repository is secure.
The approximately 100.7M-parameter causal LM began with beau-warren/blueteam-cyber-base-v0.3.0. Successive local supervised fine-tuning experiments taught defensive reporting and Open Interpreter interaction. This release branches from the strongest earlier Open Interpreter checkpoint and adds a focused one-tool contract:
{"action":"scan_repository","path":"/workspace/src"}
After receiving a bounded tool result, it returns a short overview plus exactly one indexed explanation per finding. The controller validates both stages and falls back to deterministic prose if the second response is malformed.
Only the Q8_0 GGUF inference artifact is published here. Optimizer state, training checkpoints, datasets, W&B files, credentials, and private run artifacts are excluded.
| File | Purpose |
|---|---|
blueteam-lilHybrid-cyber-LM-v0.5-Q8_0.gguf | 107,838,528-byte Q8_0 model with embedded tokenizer |
Modelfile | Tested Ollama prompt template and deterministic generation settings |
eval_results.json | Machine-readable evaluation summary |
LICENSE | MIT license |
GGUF SHA-256:
3a56cf5917ff76d9a75cf3252d3af84f5f1b313d4e6ea2f091815ba03151984f
Clone the harness repository:
git clone https://github.com/beau-warren/Cyber_Blueteam_OpenInterp_fine-tuning.git
cd Cyber_Blueteam_OpenInterp_fine-tuning
python3 -m venv .venv
.venv/bin/pip install --upgrade pip
.venv/bin/pip install -e '.[train,dev]'
make download-model
Run the included fixture:
make harness-setup
make harness-chat
Or mount your own repository read-only:
WORKSPACE=/absolute/path/to/repository make harness-setup
WORKSPACE=/absolute/path/to/repository make harness-chat
Example prompt:
Inspect /workspace for vulnerabilities, please.
Stop the stack:
WORKSPACE=/absolute/path/to/repository make harness-down
The container disables general shell/tool execution, mounts the requested repository read-only, uses an internal Docker network, drops Linux capabilities, and does not publish a host port. Review the isolation settings before mounting sensitive code.
Download the inference files:
hf download beau-warren/blueteam-lilHybrid-cyber-LM-v0.5 \
blueteam-lilHybrid-cyber-LM-v0.5-Q8_0.gguf \
Modelfile \
--local-dir blueteam-lilHybrid-cyber-LM-v0.5
cd blueteam-lilHybrid-cyber-LM-v0.5
ollama create blueteam-lilhybrid-cyber-lm-v0.5:latest -f Modelfile
ollama run blueteam-lilhybrid-cyber-lm-v0.5:latest
Direct Ollama use provides only the language model. It cannot read or scan local files without the external controller. It may emit the learned action JSON rather than a conversational answer; use the GitHub harness for actual directory scans.
The focused run used assistant-masked causal supervised fine-tuning. System, user, and tool-result tokens were context; losses applied only to assistant actions/presentations and EOS.
| Setting | Value |
|---|---|
| Parameters | 100,682,496, all trainable |
| Epochs / optimizer steps | 1 / 238 |
| Maximum sequence length | 2,048 |
| Micro-batch / accumulation / effective batch | 4 / 8 / 32 |
| Peak learning rate | 3e-6 |
| Schedule / warmup | cosine / 3% |
| Precision / optimizer | FP16 + TF32 / fused AdamW |
| Input tokens seen | 5,765,696 |
| Train / validation / test loss | 1.5614 / 1.1821 / 1.0843 |
| Training wall time | 674 seconds on an RTX 3060 Laptop GPU |
The audited SFT corpus contained complete LM→tool→LM trajectories, first-action drills, invalid and missing path cases, non-scan responses, plus frozen English/security replay. The generated corpus had zero duplicate IDs/conversations, family/template leakage, held-out harness-canary leakage, or supervised unsafe tools.
The frozen native gate contains 720 one-tool episodes decoded greedily in FP16. The parent emitted the prior report contract and scored 0% under the new one-tool protocol.
| Metric | Parent | This model |
|---|---|---|
| Valid first action | 0.00% | 99.86% |
| Correct first action | 0.00% | 82.64% |
| Exact path among scan cases | 0.00% | 85.71% |
| Valid final presentation | 0.00% | 84.82% |
| Complete episode success | 0.00% | 73.61% |
| Explicit absolute-path episodes | 0.00% | 100.00% |
| Current-workspace episodes | 0.00% | 95.63% |
| Relative-path episodes | 0.00% | 74.38% |
Q8_0/Open Interpreter gates:
/workspace selection and scan: passsrc/payments selection and scan: pass/workspace rejected without scanning: passCWE-89: passLegacy regression:
| Metric | Result |
|---|---|
| Instruction checks | 6/8 |
| Security verdict accuracy | 98.28% |
| Vulnerability precision / recall / F1 | 100.00% / 97.21% / 98.58% |
| Grounded-evidence rate | 98.28% |
| Paired consistency | 92.86% |
| Hallucinated paths | 0 |
These protocol and synthetic-suite scores are not estimates of real-world vulnerability-detection accuracy.
Intended:
Out of scope:
Released under the MIT License. Open Interpreter, Ollama, base-model data, and any downstream training datasets remain subject to their own licenses and terms.