StressEval - Table 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 table questions with all evidence in the table and a stressed reasoning step; higher is better.

Metric: Exact-match accuracy (%) on the table reasoning-stress split (294 instances seeded from WikiTableQuestions, all evidence in the table) 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
1Claude Sonnet 4.580.9#267
2Gemini 3 Pro75.5#146
3QwQ-32B74.8#913
4Qwen 2.5 72B58.5#1189
5Qwen 3 235B A22B54.1#633
6DeepSeek V3.250.3#425
7Seed-1.647.6#481
8Llama 3.1 70B44.9#1509
9Llama 3.1 8B42.2#5193

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=stresseval-table-reasoning-stress · How It Works · Data refreshed daily, snapshot 2026-11-22.