Improved performance on SWE-Bench
- rshariffdeen
- Jan 9
- 1 min read
AutoCodeRover is a fully automated approach for resolving GitHub issues (bug fixing and feature addition) where LLMs are combined with analysis and debugging capabilities to prioritize patch locations ultimately leading to a patch.
Our latest version resolves 37% of issues (pass@1) in SWE-bench lite and 46% of issues (pass@1) in SWE-bench verified! AutoCodeRover achieved this efficacy while being economical — each task costs less than $0.7 and is completed within 7 mins!
AutoCodeRover Agent
AutoCodeRover works in two stages:
🔎 Context retrieval: The LLM is provided with code search APIs to navigate the codebase and collect relevant context.
💊 Patch generation: The LLM tries to write a patch, based on retrieved context.
AutoCodeRover is a fully automated approach for resolving software engineering tasks. You can get an open-source version at our Github repository.
If you are interested in more technical details of this work, please check our Arxiv report at https://arxiv.org/abs/2408.02232
We’ll also be releasing more user-friendly services around AutoCodeRover. Stay tuned by:
Following us on X (https://twitter.com/autocoderover)
Joining our Discord channel (https://discord.gg/c6DAfNUngZ)
Visiting our Website (https://autocoderover.dev).
Comentários