Benchmark category · v0.1 · 2 verified questions
Multi-document
Reconciling documents that disagree — a supplement that amends terms, a notice that contradicts a prospectus — and identifying which governs.
Ranking
Accuracy on multi-document questions
| # | Model | Accuracy | Correct | |
|---|---|---|---|---|
| 1 | Claude Sonnet 5 | 100% | 2/2 | |
| 2 | GPT -5 | 100% | 2/2 | |
| 3 | Gemini 2.5 Pro | 100% | 2/2 | |
| Reference · CDSBench’s own system · not ranked | ||||
| — | Mastermind | 100% | 2/2 | |
Common failures
How models go wrong here
No failures recorded.
Across every model’s latest official run. Failure taxonomy →
Hardest questions
Most models wrong
Every model answered every question in this category correctly.
Latest
Most recently published questions
- #200 Republic of Sanderia 2024 (B)2026-09-15
- #62 Republic of Sanderia 20252026-09-15
Cases
2 cases with multi-document questions
- Republic of Sanderia 2025 Republic of Sanderia · 2025
- Republic of Sanderia 2024 (B) Republic of Sanderia · 2024
Following this category sends an email only when meaningful new multi-document questions are published, a model’s multi-document accuracy moves materially, or a research report on it is published. No newsletter.