DialToM - Retrospective (Persuasion): leaderboard

Retrospective theory of mind on DialToM (2026), naturalistic human-human dialogues from persuasion dialogues (PersuasionForGood): the model must identify the mental state (belief, desire, intention, emotion, knowledge or trust) of a speaker from the dialogue so far, choosing among four annotator-verified options. Correctness rate in percent; higher is better.

Metric: Correctness rate (%) on four-option multiple-choice questions of the DialToM Retrospective task: identify a speaker's mental state (belief, desire, intention, emotion, knowledge, trust) from the dialogue context, persuasion domain (naturalistic human-human dialogues from PersuasionForGood; GPT-4o-drafted options verified by professional annotators and Dawid-Skene aggregation), zero-shot; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1DialToM DeepSeek v3 (checkpoint unspecified)98.82
2Gemini 3 Pro (Preview)98.78#119
3Qwen 3 235B A22B97.94#628
4GPT-595.28#178
5Gemini 2.5 Pro94.99#276
6Gemini 2.5 Flash94.4#463
7GPT-4.194.4#505
8Kimi K293.22#443
9Llama 3.3 70B87.91#1514
10GPT-OSS-120B85.25#704
11Llama 4 Maverick79.28#1110
12Mistral Small 3.274.34#1641
13Mistral Nemo56.64#4484
14Llama 3.1 8B34.51#5196

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

Interactive version: theaggregate.ai/benchmark?slug=dialtom-retrospective-persuasion · How It Works · Data refreshed daily, snapshot 2026-11-21.