QEncodeBench - Latin Square: leaderboard
The Latin-square family of QEncodeBench (Che, Wu, Yuan and Yu, 2026): 70 core instances whose phase oracle must mark completions of a partial Latin square with distinct row and column cells. Semantic pass@1 in percent, decided by exhaustive simulation of the generated Qiskit circuit up to a global phase; higher is better.
Metric: Semantic pass@1 (%; L3 gate, Latin square completion (row and column cells distinct), 70 instances of the core set; the model writes a Qiskit build_oracle function in one shot and the verifier decides full solution-set equivalence up to a global phase by exhaustive simulation, with ancillas restored and a measurement-free unitary circuit; pass@1, one attempt per instance (five samples at temperature 0.7 for the non-reasoning DeepSeek-V4-Flash row)). Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-OSS-120B (High) | 90 | #699 (GPT-OSS-120B) |
| 2 | GPT-5.6 Terra | 85.7 | #60 |
| 3 | GPT-OSS-20B (High) | 82.9 | #1150 (GPT-OSS-20B) |
| 4 | Claude Haiku 4.5 | 81.4 | #568 |
| 5 | Claude Opus 4.8 | 77.1 | #56 |
| 6 | DeepSeek V4 Flash (Non-reasoning) | 19.1 | #254 (DeepSeek V4 Flash) |
| 7 | DeepSeek R1 0528 Qwen3 8B | 1.4 | #2101 |
| 8 | Qwen 2.5 Coder 7B Instruct | 0 | #3884 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=qencodebench-latin-square · How It Works · Data refreshed daily, snapshot 2026-11-12.