Research map / Architectures and efficiency / RNNs and normalization
Coda: An End-to-End Neural Program Decompiler
Cheng Fu, Huili Chen, Haolan Liu, Xinyun Chen, Yuandong Tian, Farinaz Koushanfar, Jishen Zhao
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add attention feeding (ablation) | Inst2AST · NES | token accuracy (%): 55.59 -> 94.66 | Improved |
| replace sequence decoder with AST tree decoder (ablation) | Inst2seq+Attn · NEL | token accuracy (%): 85.97 -> 91.92 | Improved |
| replace sequence encoder with instruction-type-aware encoder (ablation) | Seq2AST+Attn · NEL | token accuracy (%): 85.92 -> 91.92 | Improved |
| use Inst2AST+Attn instead of Seq2Seq+Attn (ablation) | Coda sketch · (Math+NE)L | token accuracy (%): 81.56 -> 93.20 | Improved |
| add iterative error correction (ablation) | Inst2AST+Attn · NEL | program accuracy (%): 2.6 -> 56.4 | Improved |
| add iterative error correction (ablation) | Inst2AST+Attn · MathS | program accuracy (%): 64.8 -> 100.0 | Improved |
| add iterative error correction (ablation) | Seq2Seq+Attn · NES | program accuracy (%): 6.6 -> 53.0 | Improved |
| increase number of ensembled error predictors from 1 to 10 (ablation) | Inst2AST EP · NES | first error detection rate (%): 67.6 -> 88.7 | Improved |
| use Inst2AST sketch generator for the error predictor instead of Seq2Seq (ablation) | EP · NES | first error detection rate (%): 83.5 -> 88.7 | Improved |
| use Coda vs RetDec/Seq2Seq-based decompiler of Katz et al. (ablation) | Coda · Math+NE | program accuracy (%): 0% -> 82% | Improved |
Closest papers by meaning
- ASAP: Assembly-Source Aligned Pseudocode Refinement For Binary Decompilation (NeurIPS 2026 · similarity 0.84)
- MOSAIC-Bench: Measuring Compositional Vulnerability Induction in Coding Agents (NeurIPS 2026 · similarity 0.84)
- Assemblage: Automatic Binary Dataset Construction for Machine Learning (NeurIPS 2024 · similarity 0.84)
- Source Code Foundation Models are Transferable Binary Analysis Knowledge Bases (NeurIPS 2024 · similarity 0.83)
- Decompile-Bench: Million-Scale Binary-Source Function Pairs for Real-World Binary Decompilation (NeurIPS 2025 · similarity 0.83)
- Instruction Tuning for Secure Code Generation (ICML 2024 · similarity 0.82)
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.
