Midm-2.0-Base-Instruct: benchmark results
Provider: KT. Released 2025-07-04. Access: Open.
Aggregate Rating 1522 ± 26, rank #1921 of 10800 rated models, from 47 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| TriBench-Ko - Omission | 61.4 | Macro-F1 (0-100) on the TriBench-Ko Omission risk items (out | 100 |
| TriBench-Ko - Adjudicative Overreach | 84.4 | Macro-F1 (0-100) on the TriBench-Ko Adjudicative Overreach r | 83.3 |
| Polar (US) - Sociocultural Axis | 84.25 | Sociocultural-axis ICAT (0-100), the StereoSet idealized con | 75.7 |
| Horangi 4 - KorQuAD 1.0 | 85.15 | F1 (x100) | 73.1 |
| Horangi 4 - KoBBQ | 92 | Accuracy (%) | 72.6 |
| Polar (South Korea) - Economic Axis | 87.24 | Economic-axis ICAT (0-100), the StereoSet idealized context | 67.6 |
| TriBench-Ko - Hallucination | 83.3 | Macro-F1 (0-100) on the TriBench-Ko Hallucination risk items | 66.7 |
| TriBench-Ko - Nondeterminism | 77.6 | Macro-F1 (0-100) on the TriBench-Ko Nondeterminism risk item | 66.7 |
| TriBench-Ko - Prompt Sensitivity | 78.2 | Macro-F1 (0-100) on the TriBench-Ko Prompt Sensitivity risk | 66.7 |
| Polar (US) | 76.12 | Total ICAT (0-100), the StereoSet idealized context associat | 64.9 |
| Polar (South Korea) | 88.17 | Total ICAT (0-100), the StereoSet idealized context associat | 59.5 |
| TriBench-Ko - Overcompliance | 66.8 | Macro-F1 (0-100) on the TriBench-Ko Overcompliance risk item | 58.3 |
Interactive version: theaggregate.ai/model?slug=midm-2-0-base-instruct · How It Works · Data refreshed daily, snapshot 2026-11-13.