0

Princeton NLP Group

Princeton University's NLP research group - creators of SWE-bench, the dominant agentic-coding benchmark.

Visit
Type
university lab
HQ
Princeton, NJ, USA
Founded
2018
Funding
University-funded; faculty grants from NSF, OpenAI, Microsoft, Open Philanthropy
Website
nlp.cs.princeton.edu
GitHub
github.com/princeton-nlp
Twitter
twitter.com/PrincetonPLI

Cite

Notes

Only stored in your browser.

Evals
1
Tools
1
Models
0
Papers
2
Boards
1
People
10

Leaderboards

1
SWE-bench LeaderboardPrinceton's canonical leaderboard for SWE-bench, SWE-bench Verified, SWE-bench Lite, and SWE-bench Multimodal, ranking coding agents by test-pass rate on real GitHub issues.

Evals

1

SWE-bench

2,294 real GitHub issues from 12 popular Python repos that require an agent to produce a patch passing the project's test suite.

ActiveCode EditingDebuggingTool CallingCode

Tools

1

Mini-SWE-Agent

~100-line minimalist SWE-agent scaffold from the original SWE-Agent team - single `bash()` tool, no internal state, designed as a reproducible baseline and as a hackable RL training target.

ScaffoldCode EditingCode GenerationTool CallingCode

Papers

2

mini-SWE-agent: A Minimal Reference Agent for SWE-bench

blog

2025

SWE-bench: Can Language Models Resolve Real-World GitHub Issues?

ICLR

2024