AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

ShieldedCode: Learning Robust Representations for Virtual Machine Protected Code

Mingqiao Mo, Yunlong Tan, Hao Zhang, Heng Zhang, Yangfan He

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning · 8 citations (OpenAlex)

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
add contrastive training (FCL+PCL) to the DeepSeekCoder VMP model (ablation)ShieldedCode · VMP corpus ablation L1Pass@1: 18.95 -> 32.17Improved
add polymorphic generation (hierarchical attention) on top of contrastive training (ablation)ShieldedCode · VMP corpus ablation (avg)Pass@1: 21.86% to 25.17%Improved
add polymorphic generation on top of contrastive training (ablation)ShieldedCode · VMP corpus ablation (avg)Pass@10: 35.25% to 38.30%Improved
add polymorphic generation, L4 (ablation)ShieldedCode · VMP corpus ablation L4Pass@1: 14.80 -> 22.36Improved
apply ShieldedCode approaches instead of standard fine-tuningGranite 3B Code 128K · HumanEval_compile (avg L1-L4)Pass@1: 12.84 -> 17.91Improved
apply ShieldedCode approaches instead of standard fine-tuningGranite 3B Code 128K · HumanEval_compile (avg L1-L4)Pass@10: 19.41 -> 25.25Improved
ShieldedCode vs GPT-4o on L0 generationShieldedCode · HumanEval_compile L0Pass@1: 22.58 -> 26.95Improved
ShieldedCode vs jTrans (Linear Probe) on O0+L1ShieldedCode · BinaryCorp-VirtualAssembly O0+L1Recall@1: 0.333 -> 0.488Improved
ShieldedCode vs jTrans (Linear Probe) on O2+L1ShieldedCode · BinaryCorp-VirtualAssembly O2+L1Recall@1: 0.715 -> 0.309Hurt
ShieldedCode vs jTrans (Linear Probe) on O1+L1ShieldedCode · BinaryCorp-VirtualAssembly O1+L1Recall@1: 0.573 -> 0.306Hurt
ShieldedCode vs VMProtect 3.7 on manual reversal successShieldedCode · Manual reverse engineering studysuccess rate: 67% -> 17%Improved
ShieldedCode vs VMProtect under angr symbolic executionShieldedCode · angr symbolic executioncompletion rate: 31% -> 3%Improved

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