AI Paper Atlas

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

NeurIPS 2023 · Large language models › Agents, code and math · approach: Software engineering agents · 2 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 monitor-guided decoding (MGD)CodeGen-350M · DotPromptscompilation rate score@6: 52.43 -> 65.37Improved
add MGDCodeGen-6B · DotPromptscompilation rate score@6: 58.64 -> 72.28Improved
add MGDSantaCoder · DotPromptsnext identifier match score@6: 82.40 -> 88.42Improved
add MGDtext-davinci-003 · DotPromptscompilation rate score@6: 62.66 -> 74.26Improved
add MGDtext-davinci-003 · DotPromptsprefix match score@6: 38.77 -> 39.94Improved
add MGD on top of RLPG prompt augmentationSantaCoder-RLPG · DotPromptscompilation rate score@6: 66.39 -> 78.14Improved
add MGD on top of classExprTypes promptsSantaCoder-classExprTypes · DotPromptscompilation rate score@6: 64.57 -> 75.01Improved
add MGD on top of fill-in-the-middle decodingSantaCoder-FIM · DotPromptscompilation rate score@6: 68.23 -> 80.19Improved
add MGD on top of FIM with classExprTypesSantaCoder-FIM-classExprTypes · DotPromptscompilation rate score@6: 70.97 -> 80.33Improved
add RLPG prompt augmentationSantaCoder · DotPromptscompilation rate score@6: 59.97 -> 66.39Improved
use fill-in-the-middle decodingSantaCoder · DotPromptscompilation rate score@6: 59.97 -> 68.23Improved
scale model from 350M to 6B parameters without MGDCodeGen · DotPromptscompilation rate score@6: 52.43 -> 58.64Improved

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