Feudor2/hallucination_bin_detector_v5.0

Model Overview

Feudor2/hallucination_bin_detector_v5.0 is an 8 billion parameter language model, fine-tuned from the YandexGPT-5-Lite-8B-instruct base model. Its primary purpose is to detect hallucinations within generated text, as indicated by its name and fine-tuning objective.

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.
• Performance: Achieved a validation loss of 0.3185 on its evaluation set, suggesting proficiency in its hallucination detection task.

Training Details

The model was trained for 5 epochs using a learning rate of 5e-05 and an Adam optimizer. Training involved a total batch size of 64 across 8 GPUs, with a cosine learning rate scheduler and a warmup ratio of 0.05.

Intended Use Cases

While specific intended uses are not detailed in the README, its fine-tuning for "hallucination_bin_detector" implies applications in:

• Content Moderation: Identifying factually incorrect or fabricated information in AI-generated content.
• Quality Assurance: Ensuring the reliability and accuracy of outputs from other large language models.
• Research: Studying and mitigating hallucination phenomena in LLMs.