WDCD R2 In-Document Resistance: leaderboard
WDCD Round 2 — resisting in-document contradictory instructions; round score.
Metric: Score (%). Source: huggingface.co. 15 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 100 | #90 |
| 2 | Grok 4 | 100 | #181 |
| 3 | O3 | 100 | #124 |
| 4 | Gemini 3.1 Pro (Preview) | 88 | #58 |
| 5 | DeepSeek V4 Pro | 88 | #102 |
| 6 | GLM-4.6 | 88 | #252 |
| 7 | GPT-6.1 Sol | 88 | #7 |
| 8 | Gemini 2.5 Pro | 75 | #152 |
| 9 | Claude Opus 4.7 | 75 | #49 |
| 10 | GPT-6 Astra | 75 | #6 |
| 11 | GPT-5.5 | 63 | #31 |
| 12 | GPT-6 Luna | 50 | #56 |
| 13 | GPT-6 Sol | 38 | #14 |
| 14 | Qwen 3 Max | 0 | #208 |
Interactive version: theaggregate.ai/benchmark?slug=wdcd-r2-in-document-resistance · How It Works · Data refreshed daily, snapshot 2026-10-20.