DART

by sheriyuo · Agent Tool · ★ 25

About DART

DART: Disentangled Action Reasoning Tuning Unofficial open-source implementation based on the paper Reasoning and Tool-use Compete in Agentic RL: From Quantifying Interference to Disentangled Tuning Core Steps Add Vocabulary Fine-tuning Use the script to perform vocabulary fine-tuning on the model: Supervised Fine-tuning (SFT) We use ms-swift for SFT training. Start the SFT training using the provided script: Note on SFT Dataset: For fairness in evaluation, the SFT training uses a self-distilled dataset.

Quick Facts

Stars25
Forks2
LanguagePython
CategoryAgent Tool
LicenseMIT
Quality Score59.7501736526283/100
Open Issues1
Last Updated2026-05-07
Created2026-03-24
Platformspython
Est. Tokens~1355k

More Agent Tool Tools

Explore other popular agent tool tools:

View all Agent Tool tools →

Popular Python Agent Tools

Frequently Asked Questions

What is DART?

DART is Reasoning and Tool-use Compete in Agentic RL: From Quantifying Interference to Disentangled Tuning. It is categorized as a Agent Tool with 25 GitHub stars.

What programming language is DART written in?

DART is primarily written in Python.

How do I install or use DART?

You can find installation instructions and usage details in the DART GitHub repository at github.com/sheriyuo/DART. The project has 25 stars and 2 forks, indicating an active community.

What license does DART use?

DART is released under the MIT license, making it free to use and modify according to the license terms.

View on GitHub → Browse Agent Tool tools