Model Overview

Feudor2/hallucination_bin_detector_v5 is an 8 billion parameter model, fine-tuned from the yandex/YandexGPT-5-Lite-8B-instruct base model. Its primary purpose is to function as a hallucination detector, indicating its specialization in identifying and flagging generated content that deviates from factual accuracy.

Key Characteristics

• Base Model: Fine-tuned from yandex/YandexGPT-5-Lite-8B-instruct.
• Parameter Count: 8 billion parameters.
• Context Length: Supports an 8192-token context window.
• Training: Trained for 3 epochs with a learning rate of 5e-05 and a total batch size of 64 across 8 GPUs. The training concluded with a validation loss of 0.3233.

Intended Use Cases

This model is specifically designed for tasks where detecting and mitigating AI hallucinations are critical. Potential applications include:

• Content Moderation: Identifying factually incorrect or fabricated information in generated text.
• Quality Assurance: Ensuring the reliability and accuracy of LLM outputs in various applications.
• Research: Studying and analyzing the prevalence and characteristics of hallucinations in language models.