Kev Release - transfer-v4 Locked Test Accuracy: leaderboard

transfer-v4 test: six never-trained public sources (QNLI, SciQ, TweetEval-offensive, PAWS, MMLU, Emotion) plus held-out policy structures; locked, read once per candidate 656 27B: test margins optimistic per card (selection after earlier test reads) Metric: accuracy.

Metric: accuracy. Source: huggingface.co. Saturation forecast: Rough model projection: around 2026. 3 models tracked. Updated 14 Nov 2026.

Top models

#ModelScoreOverall rank
1Kev 27B v10.8963
2Kev 27B0.8887
3Kev 4B0.838

Interactive version: theaggregate.ai/benchmark?slug=kev-release-transfer-v4-locked-test-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-16.