Cronus-9B
Sub-50 ms Zero-Generation Pre-Execution Security Guardrail on Qwen3.5-9B (262,144 Native Context)
Cronus-9B is a single-forward-pass, zero-generation security and authorization guardrail built on Qwen/Qwen3.5-9B. Rather than decoding text explanations, Cronus reads single-token (A..Z) log-probabilities from a frozen FP32 readout (W0) combined with a pre-folded Stage-1 backbone adapter and two role-routed all-linear Punica LoRAs (role_decision_r16, role_scope_r8).
- Training, Export & Serving Code:
FishCodeTech/cronus(GitHub) - Base Model:
Qwen/Qwen3.5-9B
Bundle Layout
Cronus-9B/
โโโ config.json # 128x128 block-wise FP8 config (Stage-1 default pre-folded)
โโโ model-00001-of-00004.safetensors ... # 4 FP8 language_model shards
โโโ model.safetensors.index.json
โโโ tokenizer.json / tokenizer_config.json / vocab.json / merges.txt / chat_template.jinja
โโโ loras/
โ โโโ role_decision_r16/ # Residual Punica LoRA (r=16, alpha=32)
โ โโโ role_scope_r8/ # Residual Punica LoRA (r=8, alpha=16)
โโโ guard_config/
โ โโโ cronus_manifest.json # Candidate manifest and SHA-256 pins
โ โโโ calibration_result.json # Frozen decision temperature & operating threshold
โ โโโ base_model.lock.json # Qwen3.5-9B 262K base lock
โ โโโ head_schema.json # Canonical SHA-256 task-head routing contracts
โ โโโ registry.json # Question-contract registry
โ โโโ profile_native.json # Pinned sm_121 Triton FLA autotuner configs
โ โโโ W0_readout/readout.safetensors # Frozen 26x4096 FP32 A..Z readout matrix + label_ids
โโโ source_bf16/
โโโ default_stage1/ # Stage-1 default LoRA (r=8, alpha=16)
โโโ role_decision_wiseft_w040/ # Modular Stage-2 decision LoRA (WiSE-FT w=0.40)
โโโ role_scope_epoch3/ # Modular Stage-2 scope LoRA (1-SE epoch-3)
โโโ combined_loras/ # Rank-concatenated BF16 single-adapter LoRAs
โโโ default_r8/
โโโ decision_r24/
โโโ scope_r16/
Quickstart with vLLM
Download the complete bundle:
hf download Sn1waR/Cronus-9B --local-dir ./Cronus-9B
Serve with cronus-guard (FishCodeTech/cronus):
python3 -m cronus.deploy.serve_vllm --bundle-dir ./Cronus-9B --host 0.0.0.0 --port 18971
Query /evaluate:
curl -s http://127.0.0.1:18971/evaluate \
-H "Content-Type: application/json" \
-d '{
"id": "demo-001",
"state": {"tool": "bash", "args": {"cmd": "ls -la /tmp/work"}, "trusted": {"scope": "authorized workspace"}},
"question": {"type": "noul", "instructions": "Is production harm expected?", "criteria": {"true": "Harm expected.", "false": "No harm."}}
}' | jq .
Benchmarks
Overall Capabilities Comparison
| Model | Scope BA | ATBench Acc* | General Decision Acc |
|---|---|---|---|
| Cronus-9B | 88.09% | 90.40% (ROC-AUC 0.9721) |
100.00% |
| THX-02 | 93.99% | 97.00% (ROC-AUC 0.9949) |
100.00% |
| Official Jev 1.14.0 | 79.96% | 81.40% | 90.41% |
| Nimble 9B v2 | 78.44% | 84.20% | 63.37% |
*ATBench-500 Evaluation & Threshold Calibration Note:
Cronus-9Bis trained strictly as a low-false-block pre-execution tool guardrail (5.88%normal false-block rate, zero ATBench training or calibration exposure) and evaluates open-ended post-hoc full-trajectory audits (ATBench-500) zero-shot via its Stage-1W0head. Because pre-execution guardrails assign conservative risk priors on post-hoc trajectories (100.00%specificity /0false positives at an uncalibrated0.50cutpoint, corresponding to58.80%raw@0.5accuracy forCronus-9Band64.60%forTHX-02), the table reports calibrated best-threshold accuracy alongside threshold-freeROC-AUC(90.40%/0.9721forCronus-9B,97.00%/0.9949forTHX-02) for an apples-to-apples capability comparison against external baselines (Official Jev 1.14.0,Nimble 9B v2) evaluated at their native operating points.
ScopeJudge Authorization Boundary (exec_unanimous n=799)
| Model | Recall (Out-of-Scope) | Specificity (In-Scope) | Balanced Accuracy (BA) |
|---|---|---|---|
Cronus-9B (ROC-AUC 0.9737) |
81.48% | 94.69% | 88.09% |
THX-02 (ROC-AUC 0.9961) |
88.89% | 99.09% | 93.99% |
| Official Jev 1.14.0 | 70.69% | 89.24% | 79.96% |
| Nimble 9B v2 | 67.24% | 89.64% | 78.44% |
| DeepSeek Distill 9B (No Boundary Training) | 58.62% | 95.63% | 77.13% |
| Laya Multilingual 322M | 93.10% | 12.28% | 52.69% |
| Needle3 | 27.59% | 73.81% | 50.70% |
| Laya English 421M | 27.59% | 73.81% | 50.70% |
| Laya Typed 421M | 17.24% | 83.65% | 50.45% |
| Community Jev-Gemma 3 270M | 0.00% | 99.59% | 49.80% |
| Binary Random Generator | 44.83% | 49.54% | 47.19% |
| Google Gemma 3 270M IT + Jevify | 36.21% | 57.36% | 46.78% |
DEV 869 & ATBench-500 Breakdown (Cronus-9B vs THX-02)
| Metric | Cronus-9B | THX-02 |
|---|---|---|
| Physical Harm BA | 94.91% | 93.98% |
| Scope Boundary BA | 93.44% | 93.94% |
| AgentApprove BA | 98.33% | 96.67% |
| Normal False-Block Rate | 5.88% | 58.82% |
| DEV Decision NLL | 0.1331 | 0.2060 |
| ATBench-500 ROC-AUC (Threshold-Free) | 0.9721 | 0.9949 |
| ATBench-500 Calibrated Best-Threshold Acc | 90.40% (88.0% Rec / 92.8% Spec) |
97.00% |
ATBench-500 Uncalibrated @0.5 Specificity / Acc |
100.00% Spec (0 FP) / 58.80% Acc |
100.00% Spec (0 FP) / 64.60% Acc |
Single-Node HTTP Serving Latency (NVIDIA DGX Spark GB10, vLLM 0.30.0, 128x128 FP8)
| Workload | P50 Latency | P95 / Max Latency |
|---|---|---|
Warm 4-question W0/permission batch (per-question forward) |
49.80 ms | 51.41 ms (51.412 ms max) |
Cold <=512-token W0/permission single-question client |
80.40 ms | โ |
Cold <=512-token all-roles single-question client |
83.04 ms | 95.18 ms |
Warm 650-token prefix-cache hit (client / forward) |
65.18 ms / 63.63 ms | โ |
Pinned Checksums
- Base model lock (
base_model.lock.json):ce6771a6a7862390f37179d427133b24515f167d6c5a159a9acc0d073302c745 W0readout (readout.safetensors):5433b51444d95c06db35ca5aaede155699e5bb92e08247d4b7ae78d970f4940c- Stage-1
default(adapter_model.safetensors):6d55a22ff6b2b80aa61e5db56d365cb8141d01575688413939b4cd86c8ce4d2a role_decision_r16modular (guard.safetensors):7f9dbcd86b7911ad9fed50043ee3c11ee49b33d7e3a74f20507c62f223717adarole_scope_r8modular (guard.safetensors):39ae42197acb684232c0cdba5385b82d6c43ad665e2b6601d93447e954375338
License
Apache-2.0
- Downloads last month
- 35