Korabbit/Llama-2-7b-chat-hf-afr-100step-flan
TEXT GENERATIONConcurrency Cost:1Model Size:7BQuant:FP8Ctx Length:4kPublished:Nov 30, 2023License:llama2Architecture:Transformer Open Weights Cold
Korabbit/Llama-2-7b-chat-hf-afr-100step-flan is a 7 billion parameter language model based on the Llama-2-7b-chat architecture, developed by Korabbit. This model is a test of the "AFR training" approach, focusing on safe, helpful, and honest assistant-like responses. It is designed to provide socially unbiased and positive interactions, explaining when questions are incoherent rather than providing incorrect information.
Loading preview...
Model Overview
Korabbit/Llama-2-7b-chat-hf-afr-100step-flan is a 7 billion parameter language model built upon the Llama-2-7b-chat foundation. Developed by Korabbit, this model serves as a testbed for the "AFR training" methodology, aiming to produce a highly helpful, respectful, and honest assistant.
Key Capabilities
- Safe and Ethical Responses: Designed to avoid harmful, unethical, racist, sexist, toxic, dangerous, or illegal content.
- Socially Unbiased: Ensures responses are socially unbiased and positive in nature.
- Coherence Handling: Explains why a question might not make sense or be factually incoherent, rather than attempting to answer incorrectly.
- Knowledge Honesty: Refrains from sharing false information if the answer to a question is unknown.
- Instruction Following: Demonstrates the ability to follow instructions, as shown by its implementation of binary search in Python.
Good For
- Applications requiring a safe and responsible AI assistant.
- Use cases where ethical guidelines and unbiased responses are paramount.
- Scenarios needing a model that can identify and gracefully handle incoherent or unanswerable queries.
- Developers interested in exploring models trained with the "AFR training" approach for enhanced safety and helpfulness.