AI Paper Atlas

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

Toward Cybersecurity-Expert Small Language Models

Matan Levi, Daniel Ohayon, Ariel Blobstein, Ravid Sa, Ian Molloy, Yair Allouche

ICML 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks

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
use SecKnowledge 2.0 instead of original SecKnowledge (ablation)CyberPal 2.0-4B · Qwen3-4B-base, avg of 9 benchmarksaverage accuracy: 68.60 -> 72.68Improved
add expert-written task formats instead of general-purpose formats (ablation)SecKnowledge 2.0 pipeline · Qwen3-4B-base, avg of 9 benchmarksaverage accuracy: 70.04 -> 72.68Improved
add retrieval search grounding (ablation)SecKnowledge 2.0 pipeline · Qwen3-4B-base, avg of 9 benchmarksaverage accuracy: 71.11 -> 72.68Improved
swap reformatting backbone gpt-oss-120B for Llama-4-Maverick (ablation)SecKnowledge 2.0 pipeline · Qwen3-4B-base, avg of 9 benchmarksaverage accuracy: 72.68 -> 72.32No clear effect
generic reformatting of original SecKnowledge (ablation)SecKnowledge · Qwen3-4B-base, avg of 9 benchmarksaverage accuracy: 68.60 -> 70.04Improved
fine-tune from base model instead of post-trained model (ablation)CyberPal 2.0-8B · CTIBench-MCQ/RCM/SecEval avgaverage accuracy: 66.47 -> 76.01Improved
fine-tune from base instead of post-trained model (ablation)CyberPal 2.0-8B · CTIBench-RCMaccuracy: 65.10 -> 85.95Improved
fine-tune on SecKnowledge 2.0 (vs post-trained Qwen3-4B)Qwen3-4B · Average of 9 benchmarksaverage accuracy: 65.19 -> 72.68Improved
fine-tune on SecKnowledge 2.0 (vs gpt-oss-20B)gpt-oss-20B · CyberMetric 2000accuracy: 90.20 -> 89.05Hurt
fine-tune on SecKnowledge 2.0 (vs gpt-oss-20B)gpt-oss-20B · Weakness Impact Mappingaccuracy: 71.91 -> 70.77Hurt
fine-tune on SecKnowledge 2.0 (vs Qwen3-14B)Qwen3-14B · CyberMetric 2000accuracy: 89.85 -> 89.95No clear effect
fine-tune on SecKnowledge 2.0 (vs Qwen3-4B)Qwen3-4B · Threat-to-TTP mappingaccuracy: 74.00 -> 82.75Improved

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