AI Paper Atlas

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

SK2Decompile: LLM-based Two-Phase Binary Decompilation from Skeleton to Skin

Hanzhuo Tan, Weihao Li, Xiaolong Tian, Siyi Wang, Jiaming Liu, Jing Li, Yuqun Zhang

ICLR 2026 · Large language models › Agents, code and math · approach: Execution-based code verification · 13 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
split decompilation into structure recovery and identifier naming via IR (ablation)SK2Decompile SFT · HumanEvalre-executability rate (%): 54.86 -> 63.75Improved
split decompilation into structure recovery and identifier naming via IR (ablation)SK2Decompile SFT · MBPPre-executability rate (%): 47.51 -> 52.83Improved
add RL with compiler and semantic rewards to the structure recovery model (ablation)pseudo-ir · HumanEvalre-executability rate (%): 62.56 -> 68.84Improved
add RL with compiler and semantic rewards to the structure recovery model (ablation)pseudo-ir · MBPPre-executability rate (%): 47.25 -> 57.06Improved
add RL to both phases of the two-phase model (ablation)pseudo-ir-src · HumanEvalre-executability rate (%): 63.75 -> 69.00Improved
add RL to both phases of the two-phase model (ablation)pseudo-ir-src · MBPPre-executability rate (%): 52.83 -> 59.63Improved
add naming phase after RL structure recovery (ablation)pseudo-ir-rl · HumanEvalre-executability rate (%): 68.84 -> 69.00No clear effect
add naming phase after RL structure recovery (ablation)pseudo-ir-rl · MBPPre-executability rate (%): 57.06 -> 59.63Improved
add independent structure recovery phase only (IR), SFT (ablation)pseudo-src · MBPPre-executability rate (%): 47.51 -> 47.25No clear effect
add IR-based two-phase decompilationSK2Decompile vs GPT-5-mini · HumanEvalre-executability rate (%): 56.75 -> 69.00Improved
add IR-based two-phase decompilationSK2Decompile vs Idioms · GitHub2025R2I: 57.97 -> 74.99Improved
add IR-based two-phase decompilationSK2Decompile vs LLM4Decompile · MBPPre-executability rate (%): 43.05 -> 59.63Improved

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