Concept Synth INDUCTION Challenge64 Round 1
General QA · 2026-07-21
Challenge64 is a 64-instance hard FullObs subset of INDUCTION evaluated on one selected direct Round-1 formula per task before symbolic repair or simplification. Accuracy uses the fixed denominator of 64, so missing or parser-invalid responses count as incorrect; the separately released generated-holdout diagnostic is not included here.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.6 Sol | 57.8 |
| Claude Fable 5 | 40.6 |
| GPT-5.6 Terra | 37.5 |
| Muse Spark 1.1 | 32.8 |
| GPT-5.6 Luna | 23.4 |
| Grok 4 | 20.3 |
| GPT-5.4 | 18.8 |
| Grok 4.5 | 17.2 |
| GPT-5.2 | 14.1 |
| Gemini 3.5 Flash | 10.9 |
| Kimi K3 | 10.9 |
| Grok 4.1 Fast | 9.4 |
| DeepSeek-V4-Pro | 9.4 |
| Opus 4.6 | 7.8 |
| Gemini 3.6 Flash | 7.8 |
| Kimi K2.7 Code | 7.8 |
| Opus 4.8 | 6.3 |
| Gemini 3 Pro Preview | 4.7 |
| Kimi K2.6 | 4.7 |
| Sonnet 5 | 4.7 |
| Grok 4.3 | 3.1 |
| DeepSeek-V3.2 | 1.6 |
| Gemini 3.1 Pro Preview | 1.6 |
| Opus 4.5 | 0.0 |
| GPT-4o | 0.0 |
| Qwen3.7-Max | 0.0 |