ALLMTS Reproduced Levine Four Level - 50 Discordance: leaderboard
ALLMTS v1: prior Levine Four Level direct-LLM prompting reproduced with Azure GPT-5.2 low,50curatedSTCCscenarios. Scored in the original method native labelspace; no external guideline execution. Support-weightedF1; discordance includes any incorrect acuity. Nativefractions. No retired original-study model attribution. Metric: Discordance (0-1).
Metric: Discordance (0-1). Source: arxiv.org. 1 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 (low) | 0.3 | #115 (GPT-5.2) |
Interactive version: theaggregate.ai/benchmark?slug=allmts-reproduced-levine-four-level-50-discordance · How It Works · Data refreshed daily, snapshot 2026-10-27.