Continual-ARC First Sight - Private Stream: leaderboard
Continual-ARC First Sight - Private Stream: First-sight exact-match solve rate (%). Continual-ARC arXiv 2611.14505v1 Table 13: first-sight exact-match solve rate, three demonstrations and one answer. Private and public tasks, stream/held-out/all and calibrated private candidates remain separate. Eight instances per rule/task except Claude Opus 5.5 has four. Table 14 is an independent harder-track cohort, not pooled with Table 13. Requested medium effort through OpenRouter is not an inference call by this scraper; Kimi does not offer that effort level.
Metric: First-sight exact-match solve rate (%). Source: arxiv.org. Saturation forecast: Around February 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) | 97 | #5 (Claude Opus 5.5) |
| 2 | GPT-6 Luna (High) | 59 | #93 (GPT-6 Luna) |
| 3 | GPT-6 Luna (Medium) | 54 | #93 (GPT-6 Luna) |
| 4 | GPT-6 Luna (Low) | 31.5 | #93 (GPT-6 Luna) |
| 5 | GLM-5.3 Flash (Medium) | 23 | #119 (GLM-5.3 Flash) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #6 GPT-6 Astra.
Interactive version: theaggregate.ai/benchmark?slug=continual-arc-first-sight-private-stream · How It Works · Data refreshed daily, snapshot 2026-11-12.