ДалееСледующее руководство
Оценка CLIP и показатели человеческих предпочтений
Технический
РУКОВОДСТВО ПО ЯЗЫКУ ИИ
Human preference data for reinforcement learning from human feedback records which responses raters prefer under stated criteria.
Прослушайте это руководствоОзвучено ИИ · ~3 min
Pairwise comparisons are a common format, but the resulting signal reflects the prompts, instructions, raters, and aggregation method rather than one universal human preference.
Collecting preference data for reinforcement learning from human feedback typically starts by generating multiple candidate responses to the same prompt from a language model, often by sampling with different random seeds or slight variations in decoding settings. Human raters are then shown two (or sometimes more) of these candidate responses side by side and asked to choose which one better satisfies criteria such as helpfulness, honesty, and harmlessness, rather than being asked to write an ideal answer themselves, which is a slower and more expensive task than making a comparative judgment. Pairwise comparisons are one common format: raters select a preferred response under explicit criteria. This avoids requiring every rater to use a numeric scale in the same way, but it does not eliminate ambiguity or bias. The chosen response depends on the prompt, comparison set, instructions, rater pool, and available abstention or tie options. Rater training is essential: without clear, detailed guidelines about what counts as helpful versus subtly unhelpful, or safe versus overly evasive, different raters will apply inconsistent standards, and inconsistent preference labels teach the resulting reward model a blurry, unreliable notion of what people actually want. Agreement checks, where multiple raters judge the same pair independently, are used to catch prompts where the 'better' answer is genuinely ambiguous or where raters are misapplying the guidelines, and persistent low agreement usually triggers a guideline revision rather than simply replacing raters. A common misconception is that RLHF preference data reflects one single, universal notion of a good response; in practice, it reflects the specific guidelines and rater pool a company chose, and different labeling instructions or rater demographics can shift a model's resulting behavior in meaningfully different directions, which is why documenting exactly what raters were asked to prioritize matters as much as collecting the comparisons themselves.
A chatbot developer shows raters two different responses to the same user question and asks which response is more helpful and accurate, recording the choice as a preference pair for training.
An AI safety team has raters compare two responses to a sensitive prompt and pick whichever one better declines an unsafe request without being preachy, building a dataset that shapes refusal behavior.
A coding assistant team shows two candidate code completions for the same prompt and has software engineers pick the one that is more correct and idiomatic, rather than just more fluent-looking.
A summarization team asks raters to compare two summaries of the same article for accuracy and conciseness, using the choices to train a reward model that scores future summaries.
Preference collection will continue to evolve alongside reward modeling, DPO, and model-assisted feedback. A useful dataset should preserve how choices were elicited and who or what provided them, because guidelines and rater pools shape the signal. For high-impact or contested domains, report disagreement and evaluate against additional evidence rather than assuming a majority preference captures everyone’s values. More efficient comparisons still require careful sampling and review. Revisit rubrics and preferences as the product, user population, or safety expectations change periodically.
Бесплатная рассылка
Три проверенных AI-истории каждое утро буднего дня, написанные простым английским. Бесплатно навсегда, без рекламы.
Одно электронное письмо каждый будний день. Отписаться в один клик. Мы никогда не продаем и не передаем ваш адрес.
Проверьте себя
Мгновенная обратная связь по каждому ответу и общий сертификат с проверяемым идентификатором после прохождения курса.
Поддержите бесплатное образование в области искусственного интеллекта. AI Understanding — это некоммерческая организация 501(c)(3). Никакой рекламы и платного доступа. Сделать пожертвование
Human preference data for reinforcement learning from human feedback records which responses raters prefer under stated criteria. Pairwise comparisons are a common format, but the resulting signal reflects the prompts, instructions, raters, and aggregation method rather than one universal human preference.
Исследование InstructGPT собрало результаты человеческого сравнения результатов кандидатов; предпочтения зависели от подсказок и инструкций оценщика.
Постоянно низкая степень согласия обычно рассматривается как признак того, что сами рекомендации нуждаются в разъяснении, а не просто как проблема с работой оценщиков.
Модель вознаграждения учится оценивать ответы в соответствии с парами человеческих предпочтений, обычно используя парные потери, такие как цель в стиле Брэдли-Терри.
DPO uses a loss function derived to have the same optimum as the traditional approach but applies it directly to the language model, without training a standalone reward model first.
Randomizing position guards against raters unconsciously favoring whichever position is shown first, which would bias the collected preference data.
Продолжайте учиться
Другие руководства, выбранные по этой теме
ДалееСледующее руководство
Оценка CLIP и показатели человеческих предпочтений
Технический