Publish auditable Perch Lint Bench scores

README.md

@@ -91,6 +91,8 @@ The rule audit reviewed all 91 original Jev misses and 30 detected controls with
 - [`benchmark.py`](https://huggingface.co/datasets/perchscan/benchmark-results/blob/main/benchmark.py): run a model on the public pairs and calculate the metrics in these tables.
 - [`adapters/`](https://huggingface.co/datasets/perchscan/benchmark-results/tree/main/adapters): example runners for the unmodified open-source models in the tables.
 - `results.jsonl`: one aggregate report per domain, track, model, and protocol.
+- [`perch-lint-results.jsonl`](https://huggingface.co/datasets/perchscan/benchmark-results/blob/main/perch-lint-results.jsonl): combined semantic-lint metrics for public, private, and total.
+- [`predictions/perch-lint/jev-1.13.0.jsonl`](https://huggingface.co/datasets/perchscan/benchmark-results/blob/main/predictions/perch-lint/jev-1.13.0.jsonl): public Perch Lint Bench method probabilities.
 - `benchmark-costs.json`: token usage, GPU timing, rates, and per-track cost calculations.
 - `split-summary.json`: split sizes and whole-file SHA-256 values. It contains no private rows or repository list.
 - `predictions/`: public-pair probabilities for auditing the public scores. No private predictions are included.

perch-lint-results.jsonl

@@ -0,0 +1,3 @@
+{"benchmark": "perch-lint-bench", "partition": "public", "protocol": "perch-semantic-ensure-v1", "model": "jev-1.13.0", "dataset_sha256": "534f6c2abddc507c40f4597607e97e07168c544011d617b83aa39e08abb35d75", "threshold": 0.5, "metrics": {"pairs": 214, "methods": 428, "average_precision": 0.9503065382620107, "precision": 0.9404761904761905, "recall": 0.7383177570093458, "f1": 0.8272251308900523, "confusion": {"tp": 158, "fp": 10, "tn": 204, "fn": 56}, "false_positives_per_true_positive": 0.06329113924050633, "pair_order": {"correct": 206, "ties": 0, "eligible": 214}, "pair_outcomes": {"both_correct": 149, "missed_before": 55, "false_alert_after": 9, "both_wrong": 1}}}
+{"benchmark": "perch-lint-bench", "partition": "private", "protocol": "perch-semantic-ensure-v1", "model": "jev-1.13.0", "dataset_sha256": "a6e9f2e4ee8ae77bdaf23af46920d97d1ba47375dc9ce03b950cdbe95c2777f9", "threshold": 0.5, "metrics": {"pairs": 71, "methods": 142, "average_precision": 0.9434782206832935, "precision": 0.9, "recall": 0.7605633802816901, "f1": 0.8244274809160306, "confusion": {"tp": 54, "fp": 6, "tn": 65, "fn": 17}, "false_positives_per_true_positive": 0.1111111111111111, "pair_order": {"correct": 67, "ties": 1, "eligible": 71}, "pair_outcomes": {"both_correct": 48, "missed_before": 17, "false_alert_after": 6, "both_wrong": 0}}}
+{"benchmark": "perch-lint-bench", "partition": "combined", "protocol": "perch-semantic-ensure-v1", "model": "jev-1.13.0", "threshold": 0.5, "metrics": {"pairs": 285, "methods": 570, "average_precision": 0.948053212754482, "precision": 0.9298245614035088, "recall": 0.743859649122807, "f1": 0.8265107212475633, "confusion": {"tp": 212, "fp": 16, "tn": 269, "fn": 73}, "false_positives_per_true_positive": 0.07547169811320754, "pair_order": {"correct": 273, "ties": 1, "eligible": 285}, "pair_outcomes": {"both_correct": 197, "missed_before": 72, "false_alert_after": 15, "both_wrong": 1}}}