CTI reasoning on AthenaBench Mini
94.33CKT AccuracyClaude-Opus-4-5
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Claude-Opus-4-5decoding=greedy2026.02 | 94.33 | 85.5 | 71.25 | 68.39 | 53.88 | 29 | |
| Claude-sonnet-4-5decoding=greedy2026.02 | 93.22 | 78.67 | 64 | 61.69 | 45.17 | 18.67 | |
| Deepseek-V3.2decoding=greedy2026.02 | 90 | 60 | 65.5 | 29.02 | 20.5 | 13 | |
| GLM-4.7decoding=greedy2026.02 | 85.67 | 66 | 70.5 | 19.45 | 29 | 12 | |
| Qwen3-Coder-Nextdecoding=greedy2026.02 | 85 | 44 | 58.5 | 5.5 | 24.5 | 8 |