Dimension Drop
Severity 10/10
2026-W12
DeepSeek V3 Stability下跌 21.4 分
Score Comparison
| Dimension | Previous | Current | Change |
|---|---|---|---|
| Overall (v5) | 52.9 | 66.6 | +13.7 |
| Code Execution (v5) | 20.2 | 62.8 | +42.6 |
| Knowledge Synthesis (v5) | 36.4 | 44.3 | +7.9 |
| Grounding (v5) | 62.3 | 78.2 | +15.9 |
| Value | 94.0 | 99.1 | +5.1 |
| Stability | 53.4 | 32.0 | -21.4 |
| Availability | 100.0 | 100.0 | +0 |
Affected Dimensions
Stability
Run #37 · Formula v5 · Judge v6 · Benchmark v5.1 · 2026-03-22 14:26 SGT
View DeepSeek V3 Full Profile