DFIR ATT&CK Technique Classification (3-Shot): leaderboard

Metric: Micro-F1 (%; multi-label ATT&CK technique prediction for 2,076 sentences from 83 The DFIR Report incident reports, 211-technique label space; three in-context examples, no chain of thought, temperature 0; models run at Q4_K_M 4-bit quantization). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 6 models tracked.

Top models

#ModelScore
1DeepSeek-V2.5 [Q4_K_M]20
2gpt-oss-120b [Q4_K_M]19
3gemma-3-27b-it [Q4_K_M]17
4gemma-3-12b-it [Q4_K_M]13
5Llama-3.1-8B-Instruct [Q4_K_M]11
6gpt-oss-20b [Q4_K_M]3

Interactive version: theaggregate.ai/benchmark?slug=dfir-att-ck-technique-classification-3-shot · How It Works · Data refreshed daily, snapshot 2026-09-26.