byte271 · Aug 18, 2026 · Aug 18, 2026 · Aug 18, 2026 · Aug 18, 2026 · Aug 18, 2026
diff --git a/.github/workflows/tiny-dev-detector.yml b/.github/workflows/tiny-dev-detector.yml
@@ -0,0 +1,69 @@
+name: Real TinyDev Detector Evidence
+
+on:
+  workflow_dispatch:
+    inputs:
+      seed_base:
+        description: First deterministic paired generation seed
+        required: true
+        default: "401000"
+        type: string
+      max_attempts:
+        description: Deterministic retry budget for exact 64-token unique pairs
+        required: true
+        default: "16"
+        type: string
+  pull_request:
+    paths:
+      - ".github/workflows/tiny-dev-detector.yml"
+      - "fuckmark/experiments/tiny_dev_detector_evidence.py"
+      - "fuckmark/tiny_dev_detector_hf.py"
+      - "fuckmark/corpus/tiny_dev_io.py"
+      - "fuckmark/corpus/tiny_dev_generation.py"
+      - "fuckmark/tiny_dev_corpus_hf.py"
+      - "tests/test_tiny_dev_detector_evidence.py"
+      - "requirements-smoke.txt"
+
+permissions:
+  contents: read
+
+jobs:
+  real-detector-evidence:
+    runs-on: ubuntu-24.04
+    timeout-minutes: 180
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+          cache: pip
+      - name: Install FuckMark
+        run: python -m pip install -e .
+      - name: Install exact CPU Torch runtime
+        run: python -m pip install "torch==2.13.0" --index-url https://download.pytorch.org/whl/cpu
+      - name: Install maintained Transformers SynthID dependencies
+        run: python -m pip install -r requirements-smoke.txt
+      - name: Reproduce real TinyDev corpus
+        run: >-
+          python -m fuckmark.tiny_dev_corpus_hf
+          --model openai-community/gpt2
+          --model-revision 607a30d783dfa663caf39e06633721c8d4cfcd7e
+          --device cpu
+          --seed-base "${{ inputs.seed_base || '401000' }}"
+          --max-attempts "${{ inputs.max_attempts || '16' }}"
+          --temperature 0.8
+          --top-k 50
+          --top-p 0.95
+          --json artifacts/tiny-dev-corpus.json
+      - name: Calibrate and evaluate Mean and Weighted Mean
+        run: >-
+          python -m fuckmark.tiny_dev_detector_hf
+          --corpus-json artifacts/tiny-dev-corpus.json
+          --device cpu
+          --json artifacts/tiny-dev-detector-evidence.json
+      - uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: real-tiny-dev-detector-evidence
+          path: artifacts/
+          if-no-files-found: error
diff --git a/evidence/tiny_dev/real-hf-gpt2-detectors-2026-08-18.json b/evidence/tiny_dev/real-hf-gpt2-detectors-2026-08-18.json
@@ -0,0 +1,203 @@
+{
+  "algorithm_version": "real-tiny-dev-detector-evidence-v1",
+  "corpus": {
+    "artifact_hash": "42240f72e17df1a048f44197f216a0b5f30128c26e2da93afbe92820946e4561",
+    "attack_negative_count": 4,
+    "attack_positive_count": 4,
+    "calibration_negative_count": 100,
+    "manifest_hash": "693788e76bceea43e72425b98f326bc09e4840d069eb47dacb18784ec7c2e388",
+    "prompt_boundary_mode": "continuation-only",
+    "prompt_count": 104,
+    "sample_count": 208,
+    "target_length": 64,
+    "token_track": "original-generation-token-ids",
+    "watermark_config_hash": "d5918a5b92bd723b78f393d986b812f786b2855c767eb7ab60192e63d2a34086"
+  },
+  "detector_artifact": {
+    "adapter_algorithm_version": "huggingface-synthid-observation-v1",
+    "adapter_config_hash": "0d9b6e5681240db5b28123042cd2ea23ad9b5886f0c3cb80d6efbe5089cb651f",
+    "adapter_id": "huggingface-transformers-synthid",
+    "artifact_hash": "d676033f294a60a6a63fd354b1604838ff6abae2c6f8f93f847156e7ca67c219",
+    "baseline_interpretability_floor": 0.8,
+    "headline_fprs": [
+      0.05,
+      0.01
+    ],
+    "primary_fpr": 0.01,
+    "scientific_status": "DEVELOPMENT_ONLY"
+  },
+  "evidence_hash": "e83cef475c5d60be02e5694f7835ac024a1e62aa9b96cdc975e51f10b105b662",
+  "evidence_kind": "development-detector-calibration-and-pristine-baseline",
+  "families": {
+    "mean": {
+      "calibration_bundle_hash": "8951996af029f636597e991a0c701a1e16f6b3b191da831ae912dcddc0b24a78",
+      "family_hash": "627484ed3b211487a9fd16ccf71848de090342f1e3d94fd8819f2afbe779aeb7",
+      "thresholds": [
+        {
+          "achieved_calibration_fpr": 0.05,
+          "attack_negative_detected_count": 0,
+          "attack_negative_fpr": 0.0,
+          "attack_negative_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.0,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.6023646356164745
+          },
+          "baseline_status": "PASS",
+          "calibration_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.01643187918205198,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.11283491110546334
+          },
+          "evaluation_hash": "c03cab481edf692bb553304500cfbe164a615432c5befdd713e71cf951da879b",
+          "pristine_detected_count": 4,
+          "pristine_sample_count": 4,
+          "pristine_tpr": 1.0,
+          "pristine_tpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.3976353643835254,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 1.0
+          },
+          "target_fpr": 0.05,
+          "threshold_hash": "b7f9955d22432de1ddba6d753a7e26ed292f0c3b7679fa73556240ea12a979c8",
+          "threshold_value": 0.538888888888889
+        },
+        {
+          "achieved_calibration_fpr": 0.01,
+          "attack_negative_detected_count": 0,
+          "attack_negative_fpr": 0.0,
+          "attack_negative_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.0,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.6023646356164745
+          },
+          "baseline_status": "PASS",
+          "calibration_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.0002531460329774256,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.05445938539208044
+          },
+          "evaluation_hash": "2abf61653c1719a5605fb96a19487c2a96c0695635f2df1c514bd7a5af3c8acb",
+          "pristine_detected_count": 4,
+          "pristine_sample_count": 4,
+          "pristine_tpr": 1.0,
+          "pristine_tpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.3976353643835254,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 1.0
+          },
+          "target_fpr": 0.01,
+          "threshold_hash": "ee158a3a242773fa5a7093cb422dd476a431278619930888e20c3b6d59fa74b3",
+          "threshold_value": 0.5555555555555555
+        }
+      ]
+    },
+    "weighted_mean": {
+      "calibration_bundle_hash": "452a2ecd65958f2bfb968730c85c96e480af84f69e40da6a27ff020b7e39c30f",
+      "family_hash": "c8a134afacec52258ba31ec1e7f6fa945ac2d5c4ad6592f9a8bff0a5747d3dfc",
+      "thresholds": [
+        {
+          "achieved_calibration_fpr": 0.05,
+          "attack_negative_detected_count": 0,
+          "attack_negative_fpr": 0.0,
+          "attack_negative_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.0,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.6023646356164745
+          },
+          "baseline_status": "PASS",
+          "calibration_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.01643187918205198,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.11283491110546334
+          },
+          "evaluation_hash": "0730644cf85866d1dc0132da9cde8eac67cf6eeaf1dc82d2a1578a93991b1b3d",
+          "pristine_detected_count": 4,
+          "pristine_sample_count": 4,
+          "pristine_tpr": 1.0,
+          "pristine_tpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.3976353643835254,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 1.0
+          },
+          "target_fpr": 0.05,
+          "threshold_hash": "1b20028af2ec9ffd30438597b750884ebddd23fb88cfb8107f14630444ca8d7e",
+          "threshold_value": 0.5455401472350625
+        },
+        {
+          "achieved_calibration_fpr": 0.01,
+          "attack_negative_detected_count": 0,
+          "attack_negative_fpr": 0.0,
+          "attack_negative_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.0,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.6023646356164745
+          },
+          "baseline_status": "PASS",
+          "calibration_fpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.0002531460329774256,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 0.05445938539208044
+          },
+          "evaluation_hash": "c69a407c4c2b5eac94fcad34da05e9aeb1ca897492ae8badf7f98213b9b3bec4",
+          "pristine_detected_count": 4,
+          "pristine_sample_count": 4,
+          "pristine_tpr": 1.0,
+          "pristine_tpr_interval": {
+            "confidence_level": 0.95,
+            "lower": 0.3976353643835254,
+            "method": "clopper-pearson-equal-tailed",
+            "upper": 1.0
+          },
+          "target_fpr": 0.01,
+          "threshold_hash": "fb518c977ed45daa128e9530e5d55989c1b6a2072f976a9ce3ebf9953311045a",
+          "threshold_value": 0.5616883116883117
+        }
+      ]
+    }
+  },
+  "runtime": {
+    "device": "cpu",
+    "python_version": "3.12.13",
+    "runner_image": "ubuntu-24.04",
+    "sampling_table_hash": "a3e9e18ea34546849c5136cf9b06c7bfd03e50e601da14af47cd926f9e623a85",
+    "sampling_table_provenance": "torch=2.13.0+cpu;device=cpu",
+    "torch_version": "2.13.0+cpu",
+    "transformers_source_commit": "a61d5f9e4fc184cff66938ff6c521cc358b5e024",
+    "transformers_version": "5.16.0.dev0"
+  },
+  "scientific_scope": "DEV_KEYS-only TinyDev Mean/Weighted Mean fixed-FPR evidence; not M6 readiness and not confirmatory evidence",
+  "source": {
+    "artifact_id": 9307352458,
+    "artifact_name": "real-tiny-dev-detector-evidence",
+    "artifact_zip_sha256": "d1174d706aa4a0b5e5124a8e8c457423cea32bb2e4d3b3161cf546c503e282d9",
+    "corpus_json_sha256": "ee09527d06abc58e612f7157b69908d5abe06101865714ce13a1287d1f0d7664",
+    "detector_json_sha256": "086770a9d8e9e7d639bad6c04365f1a239042ae2799b729831199d704188ed99",
+    "repository": "byte271/FuckMark",
+    "source_head_commit": "4e69de04f1121c7cffcca2d9b032209006f8cfca",
+    "workflow_conclusion": "success",
+    "workflow_name": "Real TinyDev Detector Evidence",
+    "workflow_run_id": 32087405132,
+    "workflow_run_number": 1
+  },
+  "verification": {
+    "corpus_reproduced_original_sha256": true,
+    "detector_json_canonical": true,
+    "detector_top_level_hash_recomputed": true,
+    "family_hashes_recomputed": true,
+    "mean_primary_baseline_passed": true,
+    "source_workflow_succeeded": true,
+    "threshold_evaluation_hashes_recomputed": true,
+    "weighted_mean_primary_baseline_passed": true
+  }
+}