MRCR v2
MRCR v2 (Multi-Round Coreference Resolution version 2) is an enhanced version of the synthetic long-context reasoning task. It extends the original MRCR framework with improved evaluation criteria and additional complexity for testing models' ability to maintain attention and reasoning across extended contexts.
5rows
scoreprimary metric
2026-05-06sampled
Metadata
Metrics
Score, Normalized Score
| Rank | Subject | Score | Model Match | Provenance | Sampled |
|---|---|---|---|---|---|
| 1 | Gemma 4 31B | 0.66 |  Gemma 4 31B google-gemma-4-31b-it | Self-reported | 2026-05-06 |
| 2 | Gemma 4 26B-A4B | 0.44 | Gemma 4 26B A4B google-gemma-4-26b-a4b-it | Self-reported | 2026-05-06 |
| 3 | Gemma 4 E4B | 0.25 | — | Self-reported | 2026-05-06 |
| 4 | Gemma 4 E2B | 0.19 | — | Self-reported | 2026-05-06 |
| 5 | Gemini 2.5 Flash-Lite | 0.17 | Gemini 2.5 Flash Lite google-gemini-2.5-flash-lite | Self-reported | 2026-05-06 |
No matching rows.