danelcsb/daniel-lfm2-350m
The danelcsb/daniel-lfm2-350m model is a personalized LFM2-350M checkpoint, adapted with LoRA, designed as a browser-native portfolio assistant for Sangbum Daniel Choi. It is specifically trained to answer questions about Daniel using supplied verified context, distinguishing between unknown facts and unrelated requests. The model excels at providing verified-profile answers and refusing out-of-scope requests, achieving an overall behavioral evaluation of 90.0%. Its primary use is to serve as a factual assistant for a personal portfolio, ensuring accurate and context-bound responses.
Loading preview...
Overview
The danelcsb/daniel-lfm2-350m is a specialized LFM2-350M language model checkpoint, fine-tuned using LoRA for deployment as a browser-native portfolio assistant. Developed by Sangbum Daniel Choi, this model is designed to provide information exclusively about Daniel based on a curated set of verified contexts.
Key Capabilities
- Context-Bound Answering: Trained on 79 curated conversations, the model answers questions strictly from supplied verified context about Daniel.
- Scope Awareness: It can differentiate between facts unknown about Daniel and requests that are unrelated to the portfolio, refusing out-of-scope queries.
- Identity Management: The model is explicitly trained to never claim to be Daniel.
- High Accuracy: Achieves a 90.0% overall behavioral evaluation, with 91.7% accuracy on verified-profile answers and 100.0% on missing-profile facts.
Good For
- Personal Portfolio Assistants: Ideal for creating AI assistants that provide factual information about an individual's professional profile.
- Contextual Q&A Systems: Suitable for applications requiring strict adherence to a defined knowledge base and refusal of irrelevant queries.
- Demonstrating LoRA Adaptation: Serves as an example of adapting a base model with LoRA for a highly specific, personalized application.