Research map / Large language models / Agents, code and math
Monitor-Guided Decoding of Code LMs with Static Analysis of Repository Context
Lakshya Agrawal, Aditya Kanade, Navin Goyal, Shuvendu K Lahiri, Sriram Rajamani
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 monitor-guided decoding (MGD) | CodeGen-350M · DotPrompts | compilation rate score@6: 52.43 -> 65.37 | Improved |
| add MGD | CodeGen-6B · DotPrompts | compilation rate score@6: 58.64 -> 72.28 | Improved |
| add MGD | SantaCoder · DotPrompts | next identifier match score@6: 82.40 -> 88.42 | Improved |
| add MGD | text-davinci-003 · DotPrompts | compilation rate score@6: 62.66 -> 74.26 | Improved |
| add MGD | text-davinci-003 · DotPrompts | prefix match score@6: 38.77 -> 39.94 | Improved |
| add MGD on top of RLPG prompt augmentation | SantaCoder-RLPG · DotPrompts | compilation rate score@6: 66.39 -> 78.14 | Improved |
| add MGD on top of classExprTypes prompts | SantaCoder-classExprTypes · DotPrompts | compilation rate score@6: 64.57 -> 75.01 | Improved |
| add MGD on top of fill-in-the-middle decoding | SantaCoder-FIM · DotPrompts | compilation rate score@6: 68.23 -> 80.19 | Improved |
| add MGD on top of FIM with classExprTypes | SantaCoder-FIM-classExprTypes · DotPrompts | compilation rate score@6: 70.97 -> 80.33 | Improved |
| add RLPG prompt augmentation | SantaCoder · DotPrompts | compilation rate score@6: 59.97 -> 66.39 | Improved |
| use fill-in-the-middle decoding | SantaCoder · DotPrompts | compilation rate score@6: 59.97 -> 68.23 | Improved |
| scale model from 350M to 6B parameters without MGD | CodeGen · DotPrompts | compilation rate score@6: 52.43 -> 58.64 | Improved |
Closest papers by meaning
- CoRe: Benchmarking LLMs’ Code Reasoning Capabilities through Static Analysis Tasks (NeurIPS 2025 · similarity 0.87)
- LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code (ICLR 2025 · similarity 0.87)
- Repoformer: Selective Retrieval for Repository-Level Code Completion (ICML 2024 · similarity 0.87)
- ObscuraCoder: Powering Efficient Code LM Pre-Training Via Obfuscation Grounding (ICLR 2025 · similarity 0.86)
- CodeChemist: Test-Time Scaling for Low-Resource Code Generation via Functional Knowledge Transfer (ICML 2026 · similarity 0.86)
- AST-T5: Structure-Aware Pretraining for Code Generation and Understanding (ICML 2024 · similarity 0.86)
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.
