Six Navers Lab / Einsia benchmarks and the work they test
Compare the tasks, scoring and evidence behind six Einsia benchmarks, with complete Radar detail pages and original references.
Model reports
PARARULE-Plus
Synthetic multi-step deductive reasoning over natural-language facts and rules under a closed-world assumption, with about 100,000 examples at each reasoning depth from 2 to 5 (nearly 400,000 in total) over animal and people entities. An aggregate number hides the per-depth breakdown the dataset was built to expose, and a model fine-tuned on the templated training split is not comparable with one prompted zero-shot.
Evidence available for this benchmark
Numeric results have not been collected for this source record. Use the paper, code or dataset below to check the task and experiments; this does not imply that the original work contains no results.
Start with these resources
Evaluation context and evidence
- Source
- Model reports
- Publisher
- Not recorded
- Benchmark release
- 2022-07-28
- Collected
- Not recorded
- Modality
- Not recorded
These are source-reported observations, not independent reruns by this site. Check the task version, evaluation protocol, unit and reporting date before comparing models. Results from different setups are not a universal ranking.
Reported results 0
This source record has no numeric results in our collection. That does not mean a score of zero or that no public results exist. Consult the original resources above.