Викторина, связанная с руководством · Жесткий Уровень
Викторина по запоминанию моделей диффузии
Проверяет, что означает запоминание в моделях диффузии, почему дублирование обучающих данных вызывает его, и как исследователи извлекают и устраняют почти копии.
Вопрос 1 из 8
Что такое запоминание в диффузионной модели?
Предварительный просмотр каждого вопроса8 вопросы
Ответы и пояснения появляются по мере прохождения теста.
- Что такое запоминание в диффузионной модели?
- The model remembering a user's previous prompts
- The model reproducing specific training images almost exactly
- The model storing images in a database for search
- The model forgetting how to draw after training
- Почему дублированные данные улучшают запоминание?
- Duplicates are stored in higher resolution
- Duplicates are removed automatically before training
- When an image appears many times, storing it becomes the easiest way to reduce training loss on it
- Duplicates confuse the model so it outputs noise
- Что продемонстрировала статья Карлини и его коллег за 2023 год?
- Diffusion models never copy training data
- Only text models memorize data
- Memorization only happens in black and white images
- Over a hundred near-identical training images could be extracted from Stable Diffusion
- Как атака извлечения идентифицировала вероятные запомненные изображения?
- It looked for near-identical generations across many seeds for the same caption
- It read the model's weights directly as images
- It asked the model which images it had memorized
- It searched the web for similar photos
- Какое утверждение о запоминании верно?
- It is the model's normal behavior for every prompt
- It is rare but can be targeted by a determined user
- It only occurs as exact pixel copies
- It has no relevance to privacy
- Why is privacy risk sharper when fine-tuning on a small dataset such as medical scans?
- Small datasets are always public
- Medical scans are too large to train on
- Each example has more influence on the model, raising the chance of regurgitation
- Fine-tuning deletes the base model
- Which mitigation bounds the influence of any single training example but currently reduces image quality?
- Differentially private training
- Increasing the number of sampling steps
- Использование более длинных подсказок
- Training for more epochs
- Why is deduplication more effective when it removes near-duplicates, not only exact copies?
- Exact copies do not affect training
- Near-duplicates are illegal to store
- Near-duplicates make images blurrier
- Slightly edited versions of the same image still act as repeated exposure