DualEvasion - Textual Evasion: leaderboard
DualEvasion (Findings of EMNLP 2026) labels 505 earnings-call answers from 60 calls for textual evasion and vocal confidence, annotated by finance professionals. This board scores zero-shot classification of the transcript text as direct or evasive by macro F1 in percent; higher is better.
Metric: Macro F1 (%; zero-shot classification of 505 earnings-call answers from 60 calls in 2023-2025 as direct or evasive from the transcript text, against labels from finance professionals; mean of the direct and evasive class F1). Source: arxiv.org. Saturation forecast: Around February 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 87.4 | #163 |
| 2 | Gemini 3 Flash | 84.7 | #159 |
| 3 | Gemini 2.5 Flash | 75.1 | #452 |
| 4 | Llama 3.1 8B Instruct | 70 | #4277 |
| 5 | Qwen 2 7B Instruct | 64.8 | #3238 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=dualevasion-textual-evasion · How It Works · Data refreshed daily, snapshot 2026-11-12.