Evolve on Graph Five-Task Average - F1: leaderboard
Evolve on Graph v1 evaluates knowledge-graph reasoning using direct answers, chain of thought, self-consistency, ReAct, retrieval, or separately trained Solver/Proposer systems. It uses Qwen2.5-3B/7B-Instruct and Llama3.1-8B-Instruct. Training lasts 140 steps, with 5 rollouts per task and at most 15 assistant turns. Hit@1 requires an exact normalized match to any gold answer; F1 measures overlap between predicted and gold answer lists. Freebase, Wikidata, and task-specific KG-R1 training scopes remain distinct. Trained weights are not aliased to the instruction backbone. Main Avg for supervised rows pools only 3 or 2 tasks, making it incomparable to the 5-task average. These remain separate task-support populations. Metric: F1 (%).
Metric: F1 (%). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 20 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | EoG study Qwen-2.5-7B-Instruct EoG | 56.31 | |
| 2 | EoG study LLaMA-3.1-8B-Instruct EoG | 53.91 | |
| 3 | EoG study Qwen-2.5-7B-Instruct Graph Reranker | 44.26 | |
| 4 | EoG study Qwen-2.5-7B-Instruct ToG | 43.66 | |
| 5 | EoG study Qwen-2.5-7B-Instruct CoR | 42.29 | |
| 6 | EoG study Qwen-2.5-3B-Instruct EoG | 42.03 | |
| 7 | EoG study Qwen-2.5-7B-Instruct Text-Based Retrieval | 40.92 | |
| 8 | EoG study Qwen-2.5-7B-Instruct PoG | 37.75 | |
| 9 | EoG study Qwen-2.5-7B-Instruct ReAct | 31.62 | |
| 10 | EoG study LLaMA-3.1-8B-Instruct ReAct | 26.09 |
Interactive version: theaggregate.ai/benchmark?slug=evolve-on-graph-five-task-average-f1 · How It Works · Data refreshed daily, snapshot 2026-10-28.