This model is a fine-tuned version of mistralai/Ministral-8B-Instruct-2410 optimized for uncertainty calibration using the ConfTuner method, as presented in the paper ConfTuner: Training Large Language Models to Express Their Confidence Verbally, accepted by NeurIPS 2025. Large Language Models (LLMs) are increasingly deployed in high-stakes domains such as science, law, and healthcare, where accurate expressions of uncertainty are essential for reliability and trust. However, current LLMs are often observed to generate incorrect answers with high confidence, a phenomenon known as «overconfidence». Recent efforts have focused on calibrating LLMs’ verbalized confidence: i.e., their expressions of confidence in text form, such as «I am 80% confident that…». Existing approaches either rely on prompt engineering or fine-tuning with heuristically generated uncertainty estimates, both of which have limited effectiveness and generalizability. Motivated by the notion of proper scoring rules for calibration in classical machine learning models, we introduce ConfTuner, a simple and efficient fine-tuning method that introduces minimal overhead and does not require ground-truth confidence…
Модальности:
Генерация текста
Области применения:
Диалог / чат
Задача: Генерация текста
Автор: liushiliushi
Теги: mistral, peft, fine-tuning, confidence-estimation, trustworthy-ai, LLM, conversational, text-generation-inference
Лайков: 4 | Загрузок: 13
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
