CRED v0.92 - Data integrity
Science · 2026-07-20
The CRED v0.92 data-integrity chapter measures detection recall on errors in how source data are selected, transformed, merged, or represented in an empirical policy evaluation.
Top models (higher is better)
| Model | Score |
|---|---|
| Sonnet 4.6 | 100.0 |
| Sonnet 5 | 100.0 |
| GPT-5.5 | 100.0 |
| GPT-5.6 Luna | 100.0 |
| GPT-5.6 Sol | 100.0 |
| Claude Fable 5 | 100.0 |
| GPT-5.6 Terra | 100.0 |
| GPT-6 Astra | 100.0 |
| GPT-6 Luna | 100.0 |
| Opus 4.8 | 100.0 |
| DeepSeek-V4-Flash | 100.0 |
| GLM-5.2 | 100.0 |
| Gemini 3.5 Flash | 100.0 |
| GPT-5.6 Luna Pro | 100.0 |
| GLM-5.3 Flash | 100.0 |
| Gemini 3.7 Flash | 100.0 |
| Gemini 3.8 Flash | 100.0 |
| Kimi K3 | 100.0 |
| Opus 4.1 | 100.0 |
| DeepSeek-V4-Pro | 97.5 |
| GLM-5 | 97.5 |
| Kimi K2.5 | 97.5 |
| Opus 4 | 97.5 |
| Sonnet 4.5 | 97.5 |
| Opus 4.5 | 97.5 |
| Kimi K2 Thinking | 95.0 |
| Kimi K2.6 | 95.0 |
| MiniMax M3 | 92.5 |
| Ring 2.6 1T | 92.5 |
| Gemini 3 Flash Preview | 87.5 |
| Hy3 preview | 85.0 |
| Nemotron 3 Ultra 550B A55B | 85.0 |
| Gemma 4 31B IT | 85.0 |
| Qwen3.5 397B A17B | 82.5 |
| DeepSeek-R1-0528 | 80.0 |
| DeepSeek-V3.2 | 80.0 |
| gpt-oss-120b | 80.0 |
| GLM-4.5 | 75.0 |
| GLM 4.6 | 75.0 |
| GLM-4.7 | 75.0 |