Perch benchmark results
This benchmark compares TypeSafe's Jev, a System One classifier, with other classifiers on finding bugs and security issues in individual code methods. Each record contains a method before and after a transaction change, plus input adapted from a Perch scan: method source, imports, callers, callees, and static call-graph edges where available. Models score each version independently, without seeing the repair. Benign reference methods are included, so a before method is not automatically positive.
Pairs come from open-source projects in Rust, JavaScript, TypeScript, Python, Go, Java, C, C++, C#, PHP, and other languages. Bug, security, and user-written semantic lint rules are separate tracks. Public and private splits are disjoint by source repository.
Inference source
| Model | Where these results were scored |
|---|---|
| Jev | TypeSafe API |
| DiffusionGemma Jev | Beam endpoint |
| SemiF | Beam endpoint |
| CLM-v0.1-8B | Runpod H100 80 GB pod (US-NE-1) |
| Span-01 | Respan scoring API (span-01-pro) |
| Liquid d1 | Liquid decision API (d1:free) |
| TypeLLM | TypeLLM API (typellm-latest) |
Reading the results
Each line shows how many issues a model finds and what percentage of clean methods it wrongly flags as its score cutoff changes. Higher and farther left is better.
The tables use each run's recorded alert settings. A value of 20% under ‘Clean methods wrongly flagged’ means one in five clean methods received an alert. Models are ranked by recall minus that percentage. Bold marks the best displayed value in each column; cost compares priced runs captured here.
Bugs
Public — 2,220 pairs
| Model | Recall | Clean methods wrongly flagged | Recall − false-positive rate | Inference cost |
|---|---|---|---|---|
| DiffusionGemma Jev | 33.4% | 27.1% | 6.3% | $0.239 |
| Jev | 17.7% | 12.7% | 5.0% | $0.551 |
| SemiF | 43.0% | 39.7% | 3.3% | $0.239 |
| TypeLLM | 8.3% | 5.5% | 2.7% | $0.576 |
| Span-01 | 25.3% | 22.6% | 2.7% | $0.182 |
| Liquid d1 | 4.1% | 2.5% | 1.6% | Free tier |
| CLM-v0.1-8B | 81.9% | 80.9% | 1.0% | ~$0.210 |
Private — 498 pairs
| Model | Recall | Clean methods wrongly flagged | Recall − false-positive rate | Inference cost |
|---|---|---|---|---|
| DiffusionGemma Jev | 35.8% | 27.3% | 8.6% | $0.054 |
| TypeLLM | 11.3% | 5.8% | 5.6% | $0.129 |
| Jev | 20.0% | 15.6% | 4.5% | $0.123 |
| Span-01 | 28.5% | 24.3% | 4.2% | $0.040 |
| SemiF | 44.3% | 41.8% | 2.6% | $0.054 |
| Liquid d1 | 5.2% | 3.0% | 2.2% | Free tier |
| CLM-v0.1-8B | 79.5% | 80.1% | -0.6% | ~$0.047 |
Security
Public — 1,311 pairs
| Model | Recall | Clean methods wrongly flagged | Recall − false-positive rate | Inference cost |
|---|---|---|---|---|
| Jev | 34.6% | 23.6% | 11.0% | $0.507 |
| TypeLLM | 27.8% | 17.4% | 10.5% | $0.566 |
| Liquid d1 | 55.6% | 46.8% | 8.8% | Free tier |
| DiffusionGemma Jev | 33.1% | 25.8% | 7.3% | $0.208 |
| Span-01 | 21.0% | 17.2% | 3.9% | $0.251 |
| SemiF | 100.0% | 99.9% | 0.1% | $0.208 |
| CLM-v0.1-8B | 100.0% | 100.0% | 0.0% | ~$1.432 |
Private — 439 pairs
| Model | Recall | Clean methods wrongly flagged | Recall − false-positive rate | Inference cost |
|---|---|---|---|---|
| Liquid d1 | 34.2% | 29.0% | 5.2% | Free tier |
| Jev | 20.6% | 16.7% | 3.9% | $0.182 |
| DiffusionGemma Jev | 21.9% | 18.5% | 3.3% | $0.074 |
| TypeLLM | 8.8% | 6.2% | 2.5% | $0.203 |
| CLM-v0.1-8B | 100.0% | 100.0% | 0.0% | ~$0.528 |
| SemiF | 100.0% | 100.0% | 0.0% | $0.074 |
| Span-01 | 8.0% | 8.1% | -0.1% | $0.091 |
Perch Lint Bench
Public — 214 pairs
| Model | Recall | Clean methods wrongly flagged | Recall − false-positive rate | Inference cost |
|---|---|---|---|---|
| TypeLLM | 61.7% | 2.8% | 58.9% | $0.050 |
| Jev | 50.0% | 0.5% | 49.5% | $0.048 |
| Liquid d1 | 49.1% | 0.9% | 48.1% | Free tier |
| DiffusionGemma Jev | 52.8% | 5.1% | 47.7% | $0.018 |
| Span-01 | 50.0% | 11.2% | 38.8% | $0.014 |
| CLM-v0.1-8B | 38.8% | 37.4% | 1.4% | ~$0.010 |
| SemiF | 32.2% | 31.3% | 0.9% | $0.018 |
Private — 71 pairs
| Model | Recall | Clean methods wrongly flagged | Recall − false-positive rate | Inference cost |
|---|---|---|---|---|
| TypeLLM | 59.2% | 5.6% | 53.5% | $0.016 |
| Jev | 53.5% | 1.4% | 52.1% | $0.016 |
| Liquid d1 | 49.3% | 1.4% | 47.9% | Free tier |
| DiffusionGemma Jev | 53.5% | 8.5% | 45.1% | $0.006 |
| Span-01 | 45.1% | 8.5% | 36.6% | $0.005 |
| CLM-v0.1-8B | 50.7% | 49.3% | 1.4% | ~$0.003 |
| SemiF | 42.3% | 45.1% | -2.8% | $0.006 |
Bug and security runs use the question floors recorded with each result. Semantic lint rules use a 70% run-wide floor.
Span-01 uses its behavior-scoring API: the target method is the output, its saved Perch context is the input, and all applicable checks are scored as behaviors in one call. Issue scores use p_present; semantic lint uses p_absent for the rule being satisfied. p_not_observable is saved separately. This is a different model interface from Jev's questions.
Liquid d1 received the same Perch method state and all applicable questions in one API call per method. Semantic lint used the same compiled ensure rule as Jev.
Inference costs cover scoring the method sides in each table. Jev costs are estimated from API-reported input tokens at TypeSafe's published rate; for semantic lint, input-token usage was measured by replaying the same 570 requests with jev-1.13.0. Span-01 costs use API-reported input tokens at Respan's published rate. Beam costs are API-reported. CLM cost is an estimated share of a 38.4-minute H100 run costing $2.23 total, allocated by embedding count. Costs are not ranked. Liquid d1 was tested on the d1:free API tier; a paid per-token price was not published. TypeLLM cost uses API-reported input tokens at its published rate; no thinking tokens were used.
JevLike (192-byte input), Laya (512-token input), and Kev (packed-branch limit) could not score every full-context method. NanoJev was unavailable on Beam. They are excluded rather than ranked on partial results.
Files and reproduction
results-current.jsonl: current-pack aggregate results for public and private splits; no private pair data.results.jsonl: all recorded result versions, identified by run and protocol ID.predictions/current/: public method probabilities from this run. Private predictions are not published.tradeoff.py: rebuilds the public recall/false-alert curves and result tables from saved scores and labels.benchmark.pyandquestion-pack.json: reproducible current-pack scorer and frozen questions.span01.py: calls Span-01 with all applicable behaviors in one request per method, saves resumable predictions, and scores the completed dataset.liquid.py: calls Liquid d1 with the frozen Perch question pack, saves resumable predictions, and scores each track.typellm.py: scores each method independently with applicable boolean questions and saves the returned probabilities.perch-lint-bench-public: paired rules and its runnable scorer.
To run Jev with the current pack, set TYPESAFE_API_KEY or PERCH_API_KEY in your environment:
python3 benchmark.py run --domain bug --scorer jev --predictions bug-jev.jsonl --report bug-jev.json
python3 benchmark.py run --domain security --scorer jev --predictions security-jev.jsonl --report security-jev.json
To run Span-01 on the public security pairs, set RESPAN_API_KEY and run:
python3 span01.py --domain security --predictions security-span01.jsonl --report security-span01.json
For another model, use --scorer command --command 'python3 your_adapter.py' --model-revision YOUR_MODEL. The adapter receives the method state and each applicable detection question and returns its true-answer probability. The runner saves predictions for resumption and reports scores only when every method side is complete.
- Downloads last month
- 317