AI Paper Atlas

Research map / Large language models / Agents, code and math

Benchmarking LLM-Assisted Blue Teaming via Standardized Threat Hunting

Yuqiao Meng, Luoxi Tang, Feiyang Yu, Xi Li, Guanhua Yan, Ping Yang, Zhaohan Xi

ICML 2026 · Large language models › Agents, code and math · approach: Software engineering agents

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace chain-of-thought prompting with standardized operational modulesGPT-5.1 · Playbook RecommendationHit@10: 90.5 -> 92.7Improved
replace chain-of-thought prompting with standardized operational modulesGPT-5.1 · Security Control AdjustmentSimilarity: 81.5 -> 91.0Improved
replace chain-of-thought prompting with standardized operational modulesGPT-5.1 · Patch Code GenerationPass rate: 77.6 -> 88.7Improved
replace chain-of-thought prompting with standardized operational modulesGPT-5.1 · Patch Tool SuggestionHit@10: 91.4 -> 98.2Improved
replace tree-of-thought prompting with standardized operational modulesGPT-5.1 · Advisory CorrelationHit@10: 85.5 -> 93.6Improved
replace in-context learning (5 examples) with standardized operational modulesGPT-5.1 · Patch Code GenerationPass rate: 59.7 -> 88.7Improved
increase in-context examples from five to tenGPT-5.1 · Playbook RecommendationHit@10: 74.4 -> 75.8No clear effect
replace chain-of-thought prompting with standardized operational modulesGrok-4.1 · Advisory CorrelationHit@10: 68.5 -> 81.6Improved
replace chain-of-thought prompting with standardized operational modulesSevenLLM-7B · Security Control AdjustmentSimilarity: 68.4 -> 80.1Improved
replace chain-of-thought prompting with standardized operational modulesLily-Cybersecurity-7B · Patch Code GenerationPass rate: 24.5 -> 29.7Improved
replace tree-of-thought with chain-of-thought promptingGPT-5.1 · Security Control AdjustmentSimilarity: 86.9 -> 81.5Hurt
add token-level noise to 10% of inputCyberTeam pipeline · noisy threat logsperformance: less than 5% performance dropHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy