MixRea - Explicit Information Accuracy: leaderboard

MixRea (2026) tests whether LLMs attend to implicit story cues while answering an explicit hypothetical question, on 2,246 four-option questions built from the Possible Stories dataset. This board is the share of questions where the chosen option reasons correctly with the explicit question information; percent, higher is better.

Metric: Explicit information accuracy (EIA, %): share of questions on which the selected option reasons correctly with the explicit information in the hypothetical question, on the 2,246 MixRea four-option explicit-implicit reasoning questions over 673 story contexts from the Possible Stories dataset, read from which of the designed options the model selects; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Flash92.5#463
2DeepSeek V392.5#729
3DeepSeek R192.4#482
4Claude 3.7 Sonnet92.1#470
5Qwen 2.5 32B91.7#1344
6Qwen 2 72B91.6#1024
7Qwen Max91.6#1092
8GPT-4o91.5#751
9Gemini 2.5 Pro91.4#276
10Llama 3.1 405B91.1#1082
11Llama 3.1 70B90.5#1499
12Llama 3 70B90.3#1655
13Qwen 2.5 72B87.9#1188
14Gemma 3 27B87.6#1663
15Gemma 2 27B87.2#2141

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

Interactive version: theaggregate.ai/benchmark?slug=mixrea-explicit-information-accuracy · How It Works · Data refreshed daily, snapshot 2026-11-21.