Airoboros-l2-70b-2.1 Chatbot Online

Airoboros-l2-70b-2.1, an experimental language model, is here to assist you in a wide range of tasks. Built upon the llama-2 base model, this chatbot offers advanced capabilities in generating text, answering questions, and executing complex instructions. In this markdown document, we'll provide you with an in-depth understanding of what this chatbot can do and how to effectively use it.

You can test out the functionality of Airoboros 70b model on Anakin AI using our online Chatbot.

Introduction to Airoboros-l2-70b-2.1 Model

Airoboros-l2-70b-2.1 is a unique and powerful chatbot created by leveraging cutting-edge techniques in natural language processing and machine learning. This model is designed to understand and respond to a variety of prompts, making it a versatile tool for a multitude of tasks.

Please note: This model is currently labeled as version 2.1, with plans for version 2.2 to address certain issues. However, it still offers robust functionality for a wide range of applications.

Model Overview

Airoboros-l2-70b-2.1 is primarily designed for experimental RP-style instruction sets, which include two main categories:

1. RP (Role Play): This category involves multi-round chats with emotes, featuring conversations between a varying number of characters defined by cards. These conversations can be engaging and interactive, allowing you to simulate dialogues with different personas.

2. GTKM (Generate, Test, Knowledge, and Memory): GTKM is a simplified alternative to ghost attention. It begins with generating a character card, followed by several questions asked to the model (as the character), using the character system prompt. Finally, everything is synthesized into a dialogue format.

Additionally, this chatbot offers experimental support for longer, more detailed writing prompts and next-chapter generation. It also includes the "stylized_response" feature, allowing for better adherence to specified styles and content based on character cards.

Helpful Usage Tips

To make the most out of Airoboros-l2-70b-2.1, here are some helpful usage tips:

Context Obedient Question Answering

For context-obedient questions, use the following format:

BEGININPUT
BEGINCONTEXT
[key0: value0]
[key1: value1]
... other metadata ...
ENDCONTEXT
[insert your text blocks here]
ENDINPUT
BEGININSTRUCTION
[insert your instructions]
ENDINSTRUCTION

Adding "Don't make up answers if you don't know" in your instruction block ensures that the model does not generate incorrect information unrelated to the context.

Coding

For coding-related tasks, you can request complex coding instructions with multiple criteria or inline criteria. You can also add "PLAINFORMAT" at the end of your prompt to get code without additional explanations.

Agent/Function Calling

Airoboros-l2-70b-2.1 is capable of generating function and argument combinations based on input criteria. You can request functions to be called, and the model will provide the necessary parameters in JSON or YAML format.

Chain-of-Thought

When tackling complex problems, you can request several potential solutions ranked by feasibility. The model will provide multiple responses and select the most suitable one.

reWOO Style Execution Planning

This feature allows you to construct systematic plans for complex instructions that require multiple tools. The output provides a plan format, and you must implement a mechanism to parse the plan and execute the functions accordingly.

License and Usage Restrictions

The airoboros-l2-70b-2.1 models are built upon the llama-2 base model, which has a custom Meta license. Please refer to the meta-license/LICENSE.txt file for the original license provided by Meta, as well as meta-license/USE_POLICY.md and meta-license/Responsible-Use-Guide.pdf.

While the license for this model is somewhat ambiguous, it's important to note that using this model commercially may have limitations due to the terms of service associated with the OpenAI API.

Airoboros-l2-70b-2.1 in the Leaderboard

Airoboros-l2-70b-2.1 has demonstrated exceptional performance in various evaluation metrics, especially in TruthfulQA. It currently holds the first position on the leaderboard, outperforming other models by a significant margin. This high TruthfulQA score ensures that the model provides reliable and accurate answers to your questions.

Whether you need assistance with writing, coding, question answering, or complex instruction execution planning, Airoboros-l2-70b-2.1 is ready to assist you with its advanced capabilities.

Feel free to explore the model and put it to the test for your specific needs. Harness the power of Airoboros-l2-70b-2.1 to simplify your tasks and enhance your productivity.

Access Airoboros-l2-70b-2.1 on Hugging Face

Please be aware that this model may continue to evolve and improve over time. Stay tuned for updates and new releases.