LogiQAte - Obfus Blood Relation (L2): leaderboard
Kinship puzzles from LogiQAte (Borah et al., 2026) in which one relation is restated as a long multi-hop chain of relations, the harder of the two obfuscation levels. Exact-match accuracy in percent, zero-shot at temperature 0; higher is better.
Metric: Exact-match accuracy (%; kinship puzzles with one relation restated as a long multi-hop chain, zero-shot, temperature 0). Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 45 | #163 |
| 2 | O4 Mini | 41 | #354 |
| 3 | GPT-4o | 27 | #741 |
| 4 | Claude Haiku 4.5 | 27 | #568 |
| 5 | GPT-4o Mini | 23 | #1627 |
| 6 | Claude 3.7 Sonnet | 21 | #455 |
| 7 | Gemini 2.5 Pro | 19 | #262 |
| 8 | QwQ-32B | 19 | #909 |
| 9 | Gemma 3 27B (IT) | 10 | #1306 |
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-blood-relation-l2 · How It Works · Data refreshed daily, snapshot 2026-11-12.