StressEval - Text Knowledge 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 text questions that need a fact withheld from the given context; higher is better.

Metric: Exact-match accuracy (%) on the text knowledge-stress split (468 instances seeded from HotpotQA, answering needs a fact withheld from the context) 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 February 2027. 9 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 3 Pro54.3#145
2Seed-1.632.9#469
3Claude Sonnet 4.531.2#261
4Llama 3.1 70B21.4#1499
5QwQ-32B17.9#907
6Qwen 3 235B A22B16.7#628
7DeepSeek V3.215.6#419
8Llama 3.1 8B15.6#5196
9Qwen 2.5 72B14.7#1188

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

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