from Hacker News

LDB: Large Language Model Debugger via Verifying Runtime Execution Step by Step

by panqueca on 4/25/24, 4:05 PM with 1 comments

• by panqueca on 4/25/24, 4:09 PM

  HumanEval Benchmark: 95.1 @ GPT-3.5

  I wonder if it can be combined with projects like SWE-Agent to build powerful yet opensource coding agents.

  - https://paperswithcode.com/sota/code-generation-on-humaneval

  - https://github.com/princeton-nlp/SWE-agent