StressEval - Knowledge Graph Reasoning Stress: leaderboard

StressEval (2026) builds Dynamic-OneEval by synthesizing new knowledge-intensive questions from GPT-5.2 failure cases on HotpotQA, ComplexWebQuestions and Wikipedia table question sets, keeping the difficulty factor that caused each failure. This board is exact-match accuracy in percent on knowledge-graph questions whose evidence is all in a synthesized graph but whose reasoning step is stressed; higher is better.

Metric: Exact-match accuracy (%) on the knowledge-graph reasoning-stress split (347 instances seeded from ComplexWebQuestions, all evidence in a synthesized virtual graph) 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: Around 2030. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1DeepSeek V3.255#419
2Gemini 3 Pro54.2#145
3Claude Sonnet 4.553.4#261
4Qwen 2.5 72B53#1188
5Seed-1.651.8#469
6Llama 3.1 70B51.3#1499
7Qwen 3 235B A22B50.7#628
8QwQ-32B50.7#907
9Llama 3.1 8B44.1#5196

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=stresseval-knowledge-graph-reasoning-stress · How It Works · Data refreshed daily, snapshot 2026-11-21.