Conversation Labeling on Cross-agent 80 labeled sessions (val)
0.93Overall KappaCodex (OpenAI)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Codex (OpenAI)Best Ver.=v72026.03 | 0.93 | 88 | 0.84 | 0.85 | 0.83 | |
| GeminiBest Ver.=v92026.03 | 0.93 | 91 | 0.85 | 0.81 | 0.87 | |
| Claude CodeBest Ver.=v52026.03 | 0.91 | 79 | 0.73 | 0.78 | 0.88 | |
| Human-HumanBest Ver.=–2026.03 | 0.89 | — | 0.78 | 0.73 | 0.7 |