CTBench - Root Cause Localization: leaderboard
CTBench (arXiv 2608.12002v1, Table 5): expert-built telecom network troubleshooting tasks solved by agent harness and model pairs in a sandbox that exposes only permitted diagnostic commands. Intersection over union between predicted and gold affected node-object sets on root cause analysis tasks, percent. Higher is better.
Metric: IoU of affected node-object sets (%). Source: arxiv.org. Saturation forecast: Around March 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 52.9 | #33 |
| 2 | Qwen 3.7 Max | 37.59 | #85 |
| 3 | Qwen 3.7 Plus | 34.68 | #98 |
| 4 | DeepSeek V4 Pro | 29.43 | #112 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ctbench-root-cause-localization · How It Works · Data refreshed daily, snapshot 2026-11-05.