whatmodel.sh / decision / Verdict 1.4 (JevBench mini)
Verdict 1.4 (JevBench mini)
↗ share
+ contribute
Verdict 1.4 (JevBench mini) is a source-available decision implementation from Heman10x.
Catalogue identity notes
JevBench 0.1.0 mini tested configuration verdict: Encoder, option markers; base ModernBERT-base; readout GLiClass bi-encoder head, RLCD/Brier calibration. OpenJev server (razorback16/openjev), which reimplements Verdict 1.4's prompt and temperatures; Verdict ships no HTTP server Separate identity because the report does not establish equivalence to the newer Decision Index checkpoint or serving recipe. No latency imported.
$ how it compares
The decision models ranked either side of Verdict 1.4 (JevBench mini), out of 357 with a score.
| # | model | score | where |
|---|---|---|---|
| 298 |  Kodiak v0.2 1B (accuracy) | 32±1 | model files |
| 299 | JEV FLAN-T5 Large public v2 (342M) | 32±1 | model files |
| 300 | Verdict 1.4 (JevBench mini) | 32±2 | source code |
| 301 | Crivo-150M v2 | 32±1 | model files |
| 302 |  Bekko System One v0 400M | 32±1 | model files |
see all decision models · find the ones that fit your machine
$ what it's good at
tools & automation▁▃▁▃▁▂▃▄▆██▆ #224 of 250 * · 15% of its score · 2 tests
Beats DM-JEPA 1.1, Kodiak v0.2 1B (accuracy), JEV FLAN-T5 Large public v2 (342M) here, though they rank higher overall. Trails Prumo (rule-based engine), Crivo-17M, JevK5 Lite here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| AgentDecisionBench: L1 BFCL v4 + When2Call (3,991 items) | 47.2% ✎ | █·····▄·▄▄·▄ beats 1/5 | 10% |
| AgentDecisionBench: L2 DecisionGym common states (200 tasks) | 19.0% ✎ | ██···█·██··█ beats 1/5 | 4.9% |
consistency▃·▃▂·▅▃·▃▅▃█ #19 of 21 · 38% of its score · 5 tests
Beats Laya (typed-decisions) (JevBench mini), Laya (english) (JevBench mini) here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| JevBench mini 0.1.0 · Probability measure coherence (pass rate) | 19.9% ✎ | ▄·▃·▃▃▁·▃▄▁█ beats 3/20 | 21% |
| JevBench mini 0.1.0 · Representation consistency (pass rate) | 66.1% ✎ | ▃·▂·▃▂▂▃▅▅▃█ beats 3/20 | 14% |
| JevBench mini 0.1.0 · Logical coherence (pass rate) | 77.9% ✎ | ▃·▃·█▅▅██▃█▅ beats 11/20 | 2.1% |
| JevBench mini 0.1.0 · Choice-set coherence (pass rate) | 33.3% ✎ | ▃▃▄█▄▄·····▄ beats 0/20 | 1.2% |
| JevBench mini 0.1.0 · Batch independence (pass rate) | 100.0% ✎ | ▁··········█ beats 3/20 | 0.0% |
other tests1 not in a task · 46% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| JevBench mini 0.1.0 · unchanged-answer accuracy | 64.2% ✎ | ▁·▃·▂···▂▂▁█ beats 2/20 | 46% |
$ where to run it
Implementation and setup instructions ↗
Published implementation and setup documentation verified. This entry describes a decision readout or evaluated configuration, not necessarily separate trained weights. Follow the referenced version and model settings; hardware compatibility has not been verified. Source · checked 2026-10-31.
Model repository and documentation: github.com/Heman10x-NGU ↗
$ how the score was worked out · JevBench 0.1.0 mini evaluatio… 84%, AgentDecisionBench 15%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 JevBench 0.1.0 mini evaluation reports | 6 | 84% |
| AgentDecisionBench | 2 | 15% |
| prior · pull toward the middle | 2% | |
| weighted average | = 32.1 |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json