Skip to content

Improve ORBIT branding and LLM routing discoverability #18

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AIGNLAI merged 1 commit into from
Oct 8, 2026
Merged

Improve ORBIT branding and LLM routing discoverability #18

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
46 changes: 33 additions & 13 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,13 @@
<p align="center">
<img src="./resources/logo.jpg" width="600px" alt="ORBIT logo">
</p>

<img src="./resources/logo.jpg" width="600px" alt="ORBIT logo">
</p>

# ORBIT: A Unified Toolbox for LLM Routing

<p align="center">
<a href="https://doi.org/10.1007/s11704-026-61310-5">Paper · Frontiers of Computer Science (2026)</a>
</p>

<p align="center">
<a href="#introduction">Introduction</a> ·
<a href="#orbit-and-routejudge">RouteJudge</a> ·
Expand All @@ -13,13 +19,16 @@
<a href="#contribution-guidelines">Contributing</a> ·
<a href="#citing-orbit-and-routejudge">Citation</a> ·
<a href="#contact">Contact</a>
</p>
## Introduction

**ORBIT (Optimal Routing and Budgeted Inference Toolbox)** is a modular, extensible toolbox for **LLM routing**. It studies how to select the most suitable model from a heterogeneous model pool **for each query**, under practical deployment constraints such as **cost, latency, and throughput**.
</p>

## Introduction

**ORBIT: A Unified Toolbox for LLM Routing** is a modular, extensible open-source toolbox for **large language model (LLM) routing**, also known as **model routing**. It studies how to select the most suitable model from a heterogeneous model pool **for each query**, under practical deployment constraints such as **cost, latency, and throughput**.

ORBIT unifies LLM router development, cost-aware inference, text-only and multimodal routing benchmarks, and reproducible performance-cost evaluation. The repository name and URL remain **ORBIT** to preserve existing paper and project links.

<p align="center">
<img src="./resources/orbit.jpg" width="600px" alt="ORBIT overview">
<img src="./resources/orbit.jpg" width="600px" alt="ORBIT unified LLM routing and evaluation workflow">
</p>

LLM routing methods are rapidly emerging, but existing implementations are often fragmented: they use different benchmark splits, cost assumptions, evaluation scripts, and router interfaces. This makes fair comparison and reproducibility difficult.
Expand Down Expand Up @@ -258,11 +267,22 @@ If you are unsure how to adapt your method to the ORBIT interface, please open a

---

## Contribution Guidelines

We welcome pull requests that improve ORBIT's method coverage, benchmark support, documentation, and evaluation utilities.

Before submitting a PR, please try to include:
## Contribution Guidelines

We welcome pull requests that improve ORBIT's method coverage, benchmark support, documentation, and evaluation utilities.

### Project naming

Use **ORBIT: A Unified Toolbox for LLM Routing** on first mention in future papers,
project introductions, and documentation for related open-source routing methods;
use **ORBIT** thereafter. Link to the canonical repository,
[LAMDA-Model-Reuse/ORBIT](https://github.com/LAMDA-Model-Reuse/ORBIT), without renaming it.
When citing the published FCS paper, retain its original bibliographic title,
**ORBIT: An Optimal Routing and Budgeted Inference Toolbox**, and DOI as listed below.

### Pull request checklist

Before submitting a PR, please try to include:

- A concise description of the method, dataset, or utility being added.
- Configuration files needed to reproduce the experiment.
Expand Down