DiagChain - Clean Evidence: leaderboard

Metric: Correctly reconstructed reference steps (%; share of the 283 coverable reference steps of the 23 clean-evidence cases (core attack evidence only); attack-chain reconstruction from heterogeneous telemetry with the ECRAG evidence-retrieval scaffold; a step counts when its evidence is observed, used, fully attributed and correctly ordered, with no earlier failure; temperature 0). Source: arxiv.org. Saturation forecast: Around July 2028. 6 models tracked.

Top models

#ModelScore
1GPT-5.538.16
2GLM-5.2 (Non-reasoning)28.27
3GLM-5.2 (High)27.56
4DeepSeek V4 Pro (Non-reasoning)26.5

Interactive version: theaggregate.ai/benchmark?slug=diagchain-clean-evidence · How It Works · Data refreshed daily, snapshot 2026-09-29.