LogiQAte - Obfus Number Series: leaderboard
Number-series completion from LogiQAte (Borah et al., 2026) with the terms written as planet names, ASCII sums or MD5 hashes of digits under a given mapping, so the rule must be recovered through the encoding. Exact-match accuracy in percent, zero-shot at temperature 0; higher is better.
Metric: Exact-match accuracy (%; number-series completion with terms replaced by planet names, ASCII sums or MD5 hashes, zero-shot, temperature 0). Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 60 | #163 |
| 2 | O4 Mini | 57 | #354 |
| 3 | QwQ-32B | 53 | #909 |
| 4 | Claude 3.7 Sonnet | 6 | #455 |
| 5 | Claude Haiku 4.5 | 5 | #568 |
| 6 | GPT-4o | 4 | #741 |
| 7 | Gemma 3 27B (IT) | 3 | #1306 |
| 8 | Gemini 2.5 Pro | 2 | #262 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=logiqate-obfus-number-series · How It Works · Data refreshed daily, snapshot 2026-11-12.