SıradakiSonraki rehber
Bradley-Terry Ödül Modellemesi
Temeller
Dil AI KILAVUZU
A reward model maps a prompt and response to a scalar score intended to approximate preference or another objective.
Bu kılavuzu dinleyinAI tarafından anlatılan · ~2 min
It is a learned proxy, not a direct measure of human values.
In RLHF-style pipelines, a reward model is often trained on human comparisons of model outputs. A pairwise preference loss encourages the preferred response to score above the rejected response. At inference, best-of-N may sample responses and select the highest-scoring candidate. Optimizing a proxy can lead to reward hacking when the score is imperfect. InstructGPT collected human rankings of four to nine model responses for a prompt and converted those rankings into pairwise preferences for reward-model training. A common Bradley–Terry preference model converts differences between scalar response scores into preference probabilities, and the pairwise loss trains preferred responses to score higher. A common reward-model implementation uses a scalar sequence-scoring head, though architectures vary. Ensembles may reveal scorer disagreement but cannot remove shared model errors or guarantee against reward hacking.
Train from pairwise response rankings.
Use scores to rank candidate responses in best-of-N selection.
Test reward predictions on held-out comparisons.
Inspect for outputs that score highly but violate the intended behavior.
Reward models remain useful components of some alignment pipelines, but require ongoing evaluation and safeguards against optimizing unintended behaviors.
Ücretsiz bülten
Hafta içi her sabah sade İngilizce yazılmış, doğrulanmış üç yapay zeka hikayesi. Sonsuza kadar ücretsiz, reklam yok.
Hafta içi her gün bir e-posta. Tek tıklamayla aboneliğinizi iptal edin. Adresinizi asla satmayız veya paylaşmayız.
Kendinizi test edin
Her yanıtla ilgili anında geri bildirim ve bir kursu geçtiğinizde doğrulanabilir bir kimliğe sahip paylaşılabilir bir sertifika.
Ücretsiz AI eğitimini destekleyin. AI Understanding, bir 501(c)(3) kâr amacı gütmeyen kuruluştur; hiçbir zaman reklam yoktur, ödeme duvarı yoktur. Bağış yap
A reward model maps a prompt and response to a scalar score intended to approximate preference or another objective. It is a learned proxy, not a direct measure of human values.
Bir ödül modeli, bir istemi ve yanıtı, tahmin edilen kaliteyi veya insan tercihini temsil eden bir skaler sayıya eşler.
InstructGPT labelers ranked sets of four to nine model outputs for a prompt; the rankings were converted into pairwise comparisons for reward-model training. This is specific to that paper’s procedure.
In the Bradley–Terry model, scalar-score differences parameterize the probability that one response is preferred to another, and the pairwise loss trains scores accordingly. These scores are not automatically calibrated measures of quality.
Ödül hackleme, modelin, kusurlu ödül modelinin yüksek oranda değerlendirdiği ancak insanların değerlendiremeyeceği kısayollar (aşırı uzunluk veya dalkavukluk gibi) bulması durumunda gerçekleşir.
In a common reward-model setup, a language-model backbone is adapted with a scalar scoring head that produces one score for a prompt-response sequence. This describes that setup, not every reward-model architecture.
Öğrenmeye devam et
Bu konu için daha fazla rehber seçildi
SıradakiSonraki rehber
Bradley-Terry Ödül Modellemesi
Temeller