StressEval - Knowledge Graph: leaderboard
Metric: Exact-match accuracy (%) on the knowledge-graph reasoning splits (knowledge stress and reasoning stress, 779 instances seeded from ComplexWebQuestions, instance-weighted mean) of Dynamic-OneEval (StressEval instances synthesized by GPT-5.2 from its own failure cases on HotpotQA, ComplexWebQuestions and WikiTableQuestions, gated for answerability and consistency), one unified prompt template for every model; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 2.5 72B | 42.4 |
| 2 | Gemini 3 Pro | 40 |
| 3 | Llama 3.1 70B | 38.3 |
| 4 | QwQ-32B | 35 |
| 5 | DeepSeek V3.2 | 33.2 |
| 6 | Claude Sonnet 4.5 | 31.5 |
| 7 | Qwen 3 235B A22B | 29.8 |
| 8 | Llama 3.1 8B | 25.9 |
| 9 | Seed-1.6 | 24.1 |
Interactive version: theaggregate.ai/benchmark?slug=stresseval-knowledge-graph · How It Works · Data refreshed daily, snapshot 2026-10-07.