Bashkir ICL - English 4-shot: leaderboard
Bashkir news-vs-analysis binary classification,macroF1,ICL without parameter updates;instruction language and demonstrations kept separate. Primary model card binds Mistral7Bv0.3 BASE,YandexGPT5Lite8BInstruct,Phi3mini4kInstruct;all use4bitNF4. Separate fine-tuned/QLoRA/CPT and generation-judge tables are not original-checkpoint scores. Macro F1 (0–1).
Metric: Macro F1 (0–1). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | YandexGPT-5-Lite-8B-instruct (NF4) | 0.58 | |
| 2 | Phi-3-mini-4k-instruct (NF4) | 0.57 | |
| 3 | Mistral-7B-v0.3 (NF4) | 0.56 |
Interactive version: theaggregate.ai/benchmark?slug=bashkir-icl-english-4-shot · How It Works · Data refreshed daily, snapshot 2026-10-24.