Datasets:
Commit History
18 days ago
-
Use a solid highlight for korjev and preserve dashed comparison models
Verified signature22393b0 -
Distinguish radar models with high-contrast colors and unique dashed patterns
Verified signature76a6e67 -
Show radar latency on a reversed linear millisecond axis with labelled ticks
Verified signature23643e5 -
Show Vega in radar legend and refresh chart image reference
Verified signature4f55c4d -
Include every measured baseline in radar, including Vega and hosted JEV
Verified signature6e68301 -
Add native Vega and hosted JEV full-set accuracy, order agreement and HTTP timing comparison
Verified signature380b992 -
Replace private-dev radar stability with full public KO/EN macro order agreement
Verified signature6aa236a
19 days ago
-
Improve radar axis arrows and background contrast
Verified signature07b19c1 -
Add eight-baseline radar and align the existing comparison table
Verified signature4b39590 -
Keep dataset card focused on public baselines and dataset documentation
Verified signaturec01629b -
Overlay all nine models on a single comparison radar
Verified signatured12edb8 -
Compare all local models with matched queue-excluded latency and korjev naming
Verified signature606241f -
Add trained Korean JEV comparison and new MMLU-Pro charts
Verified signature96e955a
23 days ago
-
Complete nine-model results and document the JEV 1.13 API evaluation
Verified signatureUpdate the dataset card, dataset details and evaluation methods with hosted API results, source breakdowns, timing, probability handling and measured costs. Publish all nine configurations in a combined JSON report. Evaluation data and labels are unchanged.9b35d4e -
Add accuracy, throughput and latency Pareto chart to the dataset card
Verified signatureShow eight single-GPU configurations and JEV 1.13 as a hosted API reference. Evaluation data and labels are unchanged.bbeb90d -
Add eight-model benchmark results and simplify the dataset card
Verified signatureMove source descriptions and label-review details to detail.md. Add full-set evaluation metrics, serving methodology, source-level breakdowns and the Cygnet temperature ablation. Evaluation data and labels are unchanged.bf2f154
24 days ago
-
Publish Korean JEV Benchmark v1: 25,418 audited evaluation items
Verified signatureSingle evaluation split with source provenance, English dataset card, source licenses, and review evidence.726173c -
initial commit
Verified signatureeaa8cec