long context benchmark
MRCR
MRCR (Multi-Round Coreference Resolution) is a synthetic long-context reasoning task where models must navigate long conversations to reproduce specific model outputs. It tests the ability to distinguish between similar requests and reason about ordering while maintaining attention across extended contexts.
Updated Aug 11, 2026
Models7
Model coverage7
MetricScore
EvidenceB
MRCR Ranking
Higher score ranks better on this benchmark.
| 01 | GO | 93.0% | 100.0% | 7 | C | |
| 02 | GO | 82.6% | 83.3% | 7 | C | |
| 03 | GO | 71.9% | 66.7% | 7 | C | |
| 04 | GO | 69.2% | 50.0% | 7 | C | |
| 05 | GO | 54.7% | 33.3% | 7 | C | |
| 06 | XI | 45.7% | 16.7% | 7 | C | |
| 07 | GO | 32.0% | 0.0% | 7 | C |
MRCR Score Distribution
A closer view of the leading scores on this benchmark.
MRCR Highlights
The leading models and scores on this benchmark.
What is MRCR?
What MRCR measures and how its scores work.
MRCR (Multi-Round Coreference Resolution) is a synthetic long-context reasoning task where models must navigate long conversations to reproduce specific model outputs. It tests the ability to distinguish between similar requests and reason about ordering while maintaining attention across extended contexts.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- MRCR
- Modality
- text
- Primary category
- long context
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- mrcr|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MRCR.
Which model scores highest on MRCR?
Gemini 2.5 Pro is currently ranked first with 93.0%.
What does MRCR measure?
MRCR (Multi-Round Coreference Resolution) is a synthetic long-context reasoning task where models must navigate long conversations to reproduce specific model outputs. It tests the ability to distinguish between similar requests and reason about ordering while maintaining attention across extended contexts.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
7 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.