Search code, repositories, users, issues, pull requests...

Clear

Search syntax tips

Provide feedback

We read every piece of feedback, and take your input very seriously.

Include my email address so I can be contacted

Cancel

Submit feedback

Saved searches

Use saved searches to filter your results more quickly

Name

Query

To see all available qualifiers, see our documentation.

Cancel

Create saved search

[Sign in](https://github.com/login?returnto=https%3A%2F%2Fgithub.com%2Fakshai0296%2Fcoderdebugger)

[Sign up](https://github.com/signup?refcta=Sign+up&refloc=header+logged+out&refpage=%2F%3Cuser-name%3E%2F%3Crepo-name%3E&source=header-repo&sourcerepo=akshai0296%2Fcoder_debugger)

Appearance settings

Resetting focus

You signed in with another tab or window. [Reload](https://github.com/akshai0296/coderdebugger) to refresh your session.You signed out in another tab or window. [Reload](https://github.com/akshai0296/coderdebugger) to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert

akshai0296/ [coder\_debugger](https://github.com/akshai0296/coder_debugger) Public

[Notifications](https://github.com/login?returnto=%2Fakshai0296%2Fcoderdebugger) You must be signed in to change notification settings
[Fork\\

0](https://github.com/login?returnto=%2Fakshai0296%2Fcoderdebugger)

[Star\\

0](https://github.com/login?returnto=%2Fakshai0296%2Fcoderdebugger)

Additional navigation options

akshai0296/coder\_debugger

main

[1 Branch](https://github.com/akshai0296/coder_debugger/branches) [0 Tags](https://github.com/akshai0296/coder_debugger/tags)

[Go to Branches page](https://github.com/akshai0296/coderdebugger/branches)[Go to Tags page](https://github.com/akshai0296/coderdebugger/tags)

Go to file

Code

Open more actions menu

Folders and files

| ## Latest commit ![akshai0296](https://github.com/akshai0296)[akshai0296](https://github.com/akshai0296/coderdebugger/commits?author=akshai0296) [Enhance README with project details and structure](https://github.com/akshai0296/coderdebugger/commit/accfa920718458f6ecc562b2a64ea9164bb4e81f) Open commit details last monthMar 13, 2026 [accfa92](https://github.com/akshai0296/coderdebugger/commit/accfa920718458f6ecc562b2a64ea9164bb4e81f) · last monthMar 13, 2026 ## History [27 Commits](https://github.com/akshai0296/coderdebugger/commits/main/) Open commit details View commit history for this file. 27 Commits |

| [Code Generation and Debugger](https://github.com/akshai0296/coderdebugger/tree/main/Code%20Generation%20and%20Debugger "Code Generation and Debugger") | [Code Generation and Debugger](https://github.com/akshai0296/coderdebugger/tree/main/Code%20Generation%20and%20Debugger "Code Generation and Debugger") | Delete Code Generation and Debugger/Fine tune | 10 months agoJun 6, 2025 |

| [Result](https://github.com/akshai0296/coderdebugger/tree/main/Result "Result") | [Result](https://github.com/akshai0296/coderdebugger/tree/main/Result "Result") | [Rename language\success\20250518\152054.png to Result/language\succes…](https://github.com/akshai0296/coderdebugger/commit/b5254b939c715af4f9fbbc628289e7a72777e088 "Rename languagesuccess20250518152054.png to Result/languagesuccess20250518_152054.png Signed-off-by: Akshai Padmanabhan <81634798+akshai0296@users.noreply.github.com>") | 10 months agoJun 6, 2025 |

| [.gitignore](https://github.com/akshai0296/coderdebugger/blob/main/.gitignore ".gitignore") | [.gitignore](https://github.com/akshai0296/coderdebugger/blob/main/.gitignore ".gitignore") | Initial commit | last yearApr 9, 2025 |

| [LICENSE](https://github.com/akshai0296/coderdebugger/blob/main/LICENSE "LICENSE") | [LICENSE](https://github.com/akshai0296/coderdebugger/blob/main/LICENSE "LICENSE") | Initial commit | last yearApr 9, 2025 |

| [README.md](https://github.com/akshai0296/coderdebugger/blob/main/README.md "README.md") | [README.md](https://github.com/akshai0296/coderdebugger/blob/main/README.md "README.md") | Enhance README with project details and structure | last monthMar 13, 2026 |

| View all files |

Repository files navigation

Coder Debugger

Permalink: Coder Debugger

A runtime-grounded evaluation and self-correction pipeline for LLM code generation across Python, Java, and C++.

Motivation

Permalink: Motivation

Large language models can generate plausible code that still fails at compile time, breaks under tests, or contains deeper logical errors. For code agents to be reliable, they need more than one-shot generation. They need a closed-loop process that can execute outputs, diagnose failures, and iteratively repair buggy solutions using runtime feedback.

This repository studies that reliability problem directly. It implements a modular generate → execute → debug pipeline and evaluates whether runtime-grounded repair improves end-to-end code generation performance.

What this project does

Permalink: What this project does

The system takes a programming task plus tests, generates candidate code with an LLM, executes the code, captures compiler traces and runtime failures, and routes the failure context to a debugger module that proposes repairs.

The current pipeline supports:

code generation
runtime execution and validation
failure capture from compiler errors, test failures, and runtime exceptions
iterative self-debugging
batch evaluation on benchmark tasks
PEFT fine-tuning with LoRA / QLoRA for an offline debugger model

Why this matters

Permalink: Why this matters

This project is aimed at reliable AI behavior rather than raw generation quality alone. It focuses on:

detecting failure modes instead of trusting surface-level plausibility
using runtime evidence as a grounding signal
measuring recovery on concrete bug classes
improving the dependability of code-generating systems

That makes it relevant to work on model evaluation, safeguards, runtime validation, and reliable agentic systems.

Pipeline overview

Permalink: Pipeline overview

The system is organized as four main components:

Generator

Produces candidate code for a task using an LLM.

Executor

Runs the generated code against tests and captures compile errors, runtime errors, and failing outputs.

Debugger

Uses the failure context to propose repairs, either through prompting or a fine-tuned offline debugger.

Coordinator

Orchestrates the full loop from generation to execution to repair.

Repository structure

Permalink: Repository structure

Plain Text

coder_debugger/
│
├── Code Generation and Debugger/
│   ├── generator/            # Code generation logic
│   │   └── generator.py
│   ├── executor/             # Runtime execution module
│   │   └── executor.py
│   ├── debugger/             # Self-debugging and repair logic
│   │   └── debugger.py
│   ├── coordinator/          # End-to-end orchestration
│   │   └── pipeline.py
│
├── scripts/
│   ├── run_pipeline.py       # Single-example execution
│   ├── batch_eval.py         # Batch evaluation on benchmarks
│   └── train_lora.py         # LoRA / QLoRA fine-tuning
│
├── benchmarks/
│   ├── results/              # Output from model runs
│   └── metrics/              # pass@1, accuracy, and related metrics

About

Runtime-grounded evaluation and self-correction pipeline for LLM code generation across Python, Java, and C++.

Resources

Readme

License

MIT license

Uh oh!

There was an error while loading. Please reload this page.

Activity

Stars

stars](https://github.com/akshai0296/coder_debugger/stargazers)

Watchers

watching](https://github.com/akshai0296/coder_debugger/watchers)

Forks

forks](https://github.com/akshai0296/coder_debugger/forks)

[Report repository](https://github.com/contact/report-content?contenturl=https%3A%2F%2Fgithub.com%2Fakshai0296%2Fcoderdebugger&report=akshai0296+%28user%29)