TradeVerse - Respondent Identification (Respondent Masked): leaderboard
Part of TradeVerse (Banerjee and Dasgupta, University of Pisa, 2026), built from World Trade Organization specific trade concern meetings: the model names the member state responding to a concern from its anonymized negotiation history, with only the respondent, its demonyms and its national agencies masked. Accuracy in percent over 1,170 concerns, higher is better; zero-shot at temperature 0.
Metric: Accuracy (%; naming the responding WTO member from the anonymized negotiation history of the 1,170 specific trade concerns, with the respondent, its demonyms and national agencies replaced by a placeholder; zero-shot, temperature 0, through the Together API). Source: arxiv.org. Saturation forecast: Estimated already saturated. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GLM-5.2 | 92.86 | #138 |
| 2 | DeepSeek V4 Pro | 92.14 | #188 |
| 3 | Kimi K2.7 Code | 91.78 | #185 |
| 4 | Llama 3.3 70B Instruct | 85.89 | #1357 |
| 5 | GPT-OSS-120B | 82.63 | #700 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=tradeverse-respondent-identification-respondent-masked · How It Works · Data refreshed daily, snapshot 2026-11-13.