SelanjutnyaPanduan berikutnya
Normalisasi Panjang dalam Optimasi Preferensi
Dasar-dasar
PANDUAN Dasar
Grouped reward normalization expresses each sampled response’s reward relative to other responses for the same prompt.
Dengarkan panduan iniDiceritakan oleh AI · ~2 min
DeepSeek introduced this mechanism in Group Relative Policy Optimization (GRPO), a PPO variant that omits a separate value-model critic. It provides a relative training signal, but does not guarantee that prompt difficulty or reward scale is fully corrected.
DeepSeekMath introduced Group Relative Policy Optimization (GRPO) for language-model reinforcement learning. For each prompt, the policy samples a group of responses and computes rewards. Each response’s advantage is based on its reward relative to the group mean and, in the original outcome-supervision formulation, the group standard deviation. The group statistic replaces the separate learned value-model baseline used in PPO. DeepSeekMath applied GRPO to mathematical reasoning with verifiable rewards; the method is not limited to human-preference RLHF. When all responses receive the same reward, centering gives no relative advantage, and a zero standard deviation requires numerical handling in implementations.
Comparing several candidate math solutions from one prompt using a verifier score
Using code-test results to rank responses generated for the same coding prompt
Inspecting how a reward model’s scores vary within one prompt group
Reducing memory use in a PPO-style language-model update by omitting a separate value network
Grouped policy optimization is used in work on verifiable-reward reasoning, but its behavior depends on group sampling, reward functions, normalization choices, and implementation details. Research continues on reward diversity, process supervision, and stability. Evaluation should distinguish memory savings from improvements in reasoning quality.
Buletin gratis
Tiga cerita AI terverifikasi setiap pagi hari kerja, ditulis dalam bahasa Inggris sederhana. Gratis selamanya, tanpa iklan.
Satu email setiap hari kerja. Berhenti berlangganan dalam satu klik. Kami tidak pernah menjual atau membagikan alamat Anda.
Uji diri Anda sendiri
Umpan balik instan untuk setiap jawaban, dan sertifikat yang dapat dibagikan dengan ID yang dapat diverifikasi setelah Anda lulus kursus.
Mendukung pendidikan AI gratis. AI Understanding adalah organisasi nirlaba 501(c)(3) — tanpa iklan, tanpa paywall, selamanya. Berikan sumbangan
Grouped reward normalization expresses each sampled response’s reward relative to other responses for the same prompt. DeepSeek introduced this mechanism in Group Relative Policy Optimization (GRPO), a PPO variant that omits a separate value-model critic. It provides a relative training signal, but does not guarantee that prompt difficulty or reward scale is fully corrected.
Setiap hadiah diubah menjadi skor-z menggunakan mean dan deviasi standar dari kelompok respons terhadap perintah yang sama.
GRPO, diperkenalkan oleh DeepSeek dan digunakan di DeepSeek-R1, dibangun dengan menormalisasi imbalan dalam grup.
Dengan menggunakan rata-rata grup sebagai dasar, GRPO menghilangkan kritik yang dipelajari, sehingga menghemat memori dan komputasi.
Mengurangi rata-rata kelompok akan membuat tanggapan yang berada di bawah rata-rata mempunyai keuntungan negatif, sehingga menjauhkan kebijakan tersebut dari tanggapan mereka.
Group normalization supplies relative advantages among responses to one prompt and scales them by the group spread; it does not guarantee that all prompt-specific differences vanish.
Teruslah belajar
Panduan lainnya dipilih untuk topik ini
SelanjutnyaPanduan berikutnya
Normalisasi Panjang dalam Optimasi Preferensi
Dasar-dasar