reasoning benchmark
GraphWalks
GraphWalks is a synthetic multi-hop long-context reasoning benchmark in which a model is given an edge-list representation of a graph and must traverse it to find neighboring nodes (via breadth-first search) or parent nodes for a given start node. Performance is reported as F1 of the model-predicted answer set versus the ground truth.
Updated Aug 7, 2026
Published models3
Registry coverage3
MetricScore
EvidenceB
GraphWalks leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MI | 90.0% | 100.0% | 3 | C | |
| 2 | XI | 87.0% | 50.0% | 3 | C | |
| 3 | XI | 62.0% | 0.0% | 3 | C |
Score distribution
Top published rows on the benchmark's original scale.
GraphWalks highlights
The top published results on this benchmark's own scale.
What is GraphWalks?
Definition and scoring fields from the benchmark registry.
GraphWalks is a synthetic multi-hop long-context reasoning benchmark in which a model is given an edge-list representation of a graph and must traverse it to find neighboring nodes (via breadth-first search) or parent nodes for a given start node. Performance is reported as F1 of the model-predicted answer set versus the ground truth.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- GraphWalks
- Modality
- text
- Primary category
- reasoning
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- graphwalks|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about GraphWalks.
Which model scores highest on GraphWalks?
MAI-Thinking-1 is currently ranked first with 90.0%.
What does GraphWalks measure?
GraphWalks is a synthetic multi-hop long-context reasoning benchmark in which a model is given an edge-list representation of a graph and must traverse it to find neighboring nodes (via breadth-first search) or parent nodes for a given start node. Performance is reported as F1 of the model-predicted answer set versus the ground truth.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.