Arbor
Autonomous research agent that forms hypotheses, edits code and runs experiments, keeping only gains that hold on held-out data
Agent by RUC-NLPIR
IndexedApache-2.0★ 1,125
Get started
Setup steps are in the repository. Website · Repository
About
Arbor takes a benchmark and a goal, proposes hypotheses in a tree, edits code and runs experiments. Each executor works in its own git worktree and merges a gain only if it clears a configurable margin on a held-out test split. It also searches literature through the alphaXiv API, can check an idea's novelty before compute is spent, and ships as a CLI with an Agent Skill suite for Codex and Claude Code.
Use cases: research, coding, deep-research, code-generation, llm-evaluation
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Fact sheet
- Jobs
- research, codingREADME
- Autonomy
- Autonomous: runs tasks end to end and reports backREADME
- Where you use it
- TerminalREADME
- Hosting
- You run itREADME
- Interfaces
- agent-skillsREADME
- Open source
- YesREADME
Facts extracted by Claude Haiku 5.5 from the linked sources on Oct 24, 2026. Wrong fact? Tell us.
Signals
- Stars
- 1,125
- Forks
- 128
- Open issues
- 1
- Last commit
- Sep 8, 2026 (1 month ago)
- Last release
- Jul 12, 2026 (3 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- Apache-2.0
- Repository
- github.com/RUC-NLPIR/Arbor
- Website
- ruc-nlpir.github.io/Arbor/
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- READMEgithub-graphql
Cite this page
Arbor. Agent Mart. https://agentmart.dev/agents/arbor (last updated 2026-10-24)