Rank benchmark results by F1 and highlight metric leaders
#14
by jfbrown95 - opened
README.md
CHANGED
|
@@ -12,7 +12,7 @@ The pairs come from open-source projects in Rust, JavaScript, TypeScript, Python
|
|
| 12 |
|
| 13 |
## Reading the results
|
| 14 |
|
| 15 |
-
AP (average precision) measures ranking across score cutoffs. For the other metrics, a score of at least 0.5 counts as an alert. Precision is the share of alerts that are correct; recall is the share of issue-bearing methods found; F1 is their harmonic mean. TP is a correct alert, FP a false alert, and FN a missed issue. Pair order is the share of eligible before/after pairs where the before method scores higher than the fixed after method; ties do not count as correct.
|
| 16 |
|
| 17 |
## Bugs
|
| 18 |
|
|
@@ -22,19 +22,19 @@ AP (average precision) measures ranking across score cutoffs. For the other metr
|
|
| 22 |
|
| 23 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 24 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 25 |
-
| [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) | 47.6% | 49.9% | 22.0% | 30.6% | 410 | 412 | 1451 | 42.5% |
|
|
|
|
| 26 |
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 44.1% | 47.5% | 7.6% | 13.1% | 141 | 156 | 1720 | 32.1% |
|
| 27 |
-
| [
|
| 28 |
-
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 43.7% | 44.8% | 2.3% | 4.4% | 43 | 53 | 1818 | 22.4% |
|
| 29 |
|
| 30 |
### Private — 498 pairs
|
| 31 |
|
| 32 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 33 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 34 |
-
| [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) | 48.6% | 50.7% | 24.1% | 32.6% | 102 | 99 | 322 | 43.4% |
|
|
|
|
| 35 |
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 45.8% | 46.4% | 9.2% | 15.4% | 39 | 45 | 385 | 32.7% |
|
| 36 |
-
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 45.2% | 60.0% | 3.5% | 6.7% | 15 | 10 | 409 | 22.7% |
|
| 37 |
-
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 45.2% | 44.6% | 17.5% | 25.1% | 74 | 92 | 350 | 43.4% |
|
| 38 |
|
| 39 |
## Security
|
| 40 |
|
|
@@ -44,21 +44,21 @@ AP (average precision) measures ranking across score cutoffs. For the other metr
|
|
| 44 |
|
| 45 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 46 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 47 |
-
| [
|
| 48 |
-
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 44.8% | 46.4% | 12.4% | 19.6% | 142 | 164 | 1000 | 30.6% |
|
| 49 |
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 44.0% | 43.6% | 64.9% | 52.1% | 741 | 959 | 401 | 41.5% |
|
| 50 |
-
| [
|
| 51 |
-
| [
|
|
|
|
| 52 |
|
| 53 |
### Private — 439 pairs
|
| 54 |
|
| 55 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 56 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 57 |
-
| [
|
| 58 |
-
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 47.4% | 45.0% | 4.5% | 8.2% | 18 | 22 | 380 | 37.9% |
|
| 59 |
-
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 47.3% | 43.8% | 1.8% | 3.4% | 7 | 9 | 391 | 29.4% |
|
| 60 |
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 47.1% | 46.9% | 67.8% | 55.4% | 270 | 306 | 128 | 52.8% |
|
| 61 |
-
| [
|
|
|
|
|
|
|
| 62 |
|
| 63 |
† CLM-v0.1-8B uses its choice head with a 6,144-token encoder; SecureBERT2.0 uses its detection head with a 4,096-token input cap. All models are scored on the same method sides and labels within each track.
|
| 64 |
|
|
|
|
| 12 |
|
| 13 |
## Reading the results
|
| 14 |
|
| 15 |
+
AP (average precision) measures ranking across score cutoffs. For the other metrics, a score of at least 0.5 counts as an alert. Precision is the share of alerts that are correct; recall is the share of issue-bearing methods found; F1 is their harmonic mean. TP is a correct alert, FP a false alert, and FN a missed issue. Pair order is the share of eligible before/after pairs where the before method scores higher than the fixed after method; ties do not count as correct. Each table is ranked by F1 at 0.5. Bold marks the best value in each column (lowest for FP and FN).
|
| 16 |
|
| 17 |
## Bugs
|
| 18 |
|
|
|
|
| 22 |
|
| 23 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 24 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 25 |
+
| [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) | **47.6%** | **49.9%** | **22.0%** | **30.6%** | **410** | 412 | **1451** | 42.5% |
|
| 26 |
+
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 44.0% | 44.9% | 13.9% | 21.2% | 258 | 317 | 1603 | **46.9%** |
|
| 27 |
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 44.1% | 47.5% | 7.6% | 13.1% | 141 | 156 | 1720 | 32.1% |
|
| 28 |
+
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 43.7% | 44.8% | 2.3% | 4.4% | 43 | **53** | 1818 | 22.4% |
|
|
|
|
| 29 |
|
| 30 |
### Private — 498 pairs
|
| 31 |
|
| 32 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 33 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 34 |
+
| [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) | **48.6%** | 50.7% | **24.1%** | **32.6%** | **102** | 99 | **322** | **43.4%** |
|
| 35 |
+
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 45.2% | 44.6% | 17.5% | 25.1% | 74 | 92 | 350 | **43.4%** |
|
| 36 |
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 45.8% | 46.4% | 9.2% | 15.4% | 39 | 45 | 385 | 32.7% |
|
| 37 |
+
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 45.2% | **60.0%** | 3.5% | 6.7% | 15 | **10** | 409 | 22.7% |
|
|
|
|
| 38 |
|
| 39 |
## Security
|
| 40 |
|
|
|
|
| 44 |
|
| 45 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 46 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 47 |
+
| [SecureBERT2.0](https://huggingface.co/cisco-ai/SecureBERT2.0-code-vuln-detection)† | 43.6% | 43.5% | **99.8%** | **60.6%** | **1140** | 1479 | **2** | **50.2%** |
|
|
|
|
| 48 |
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 44.0% | 43.6% | 64.9% | 52.1% | 741 | 959 | 401 | 41.5% |
|
| 49 |
+
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 44.8% | 46.4% | 12.4% | 19.6% | 142 | 164 | 1000 | 30.6% |
|
| 50 |
+
| [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) | **49.2%** | **54.4%** | 10.2% | 17.2% | 117 | 98 | 1025 | 40.5% |
|
| 51 |
+
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 43.3% | 36.0% | 3.2% | 5.8% | 36 | **64** | 1106 | 22.9% |
|
| 52 |
|
| 53 |
### Private — 439 pairs
|
| 54 |
|
| 55 |
| Model | AP | Precision at 0.5 | Recall at 0.5 | F1 at 0.5 | TP | FP | FN | Pair order |
|
| 56 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
|
| 57 |
+
| [SecureBERT2.0](https://huggingface.co/cisco-ai/SecureBERT2.0-code-vuln-detection)† | 43.0% | 45.4% | **99.7%** | **62.4%** | **397** | 478 | **1** | **55.8%** |
|
|
|
|
|
|
|
| 58 |
| [CLM-v0.1-8B](https://huggingface.co/Contrastive-LM/CLM-v0.1-8B)† | 47.1% | 46.9% | 67.8% | 55.4% | 270 | 306 | 128 | 52.8% |
|
| 59 |
+
| [JevK5](https://huggingface.co/alibiserikbay/JevK5) | 47.4% | 45.0% | 4.5% | 8.2% | 18 | 22 | 380 | 37.9% |
|
| 60 |
+
| [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) | **48.9%** | **52.6%** | 2.5% | 4.8% | 10 | **9** | 388 | 36.2% |
|
| 61 |
+
| [jev-lite](https://huggingface.co/vagmi/jev-lite) | 47.3% | 43.8% | 1.8% | 3.4% | 7 | **9** | 391 | 29.4% |
|
| 62 |
|
| 63 |
† CLM-v0.1-8B uses its choice head with a 6,144-token encoder; SecureBERT2.0 uses its detection head with a 4,096-token input cap. All models are scored on the same method sides and labels within each track.
|
| 64 |
|