SeterusnyaPanduan seterusnya
Normalisasi Panjang dalam Pengoptimuman Keutamaan
Asas
PANDUAN Asas
Grouped reward normalization expresses each sampled response’s reward relative to other responses for the same prompt.
Dengar panduan iniAI-diceritakan · ~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.
Newsletter percuma
Tiga cerita AI yang disahkan setiap pagi hari bekerja, ditulis dalam Bahasa Inggeris biasa. Percuma selamanya, tiada iklan.
Satu e-mel setiap hari bekerja. Berhenti melanggan dalam satu klik. Kami tidak pernah menjual atau berkongsi alamat anda.
Uji diri anda
Maklum balas segera pada setiap jawapan, dan sijil yang boleh dikongsi dengan ID yang boleh disahkan sebaik sahaja anda lulus kursus.
Sokong pendidikan AI percuma. AI Understanding ialah pertubuhan bukan keuntungan 501(c)(3) — tiada iklan, tiada bayaran dinding, selamanya. Buat derma
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 ganjaran ditukar kepada skor z menggunakan min dan sisihan piawai kumpulan respons kepada gesaan yang sama.
GRPO, yang diperkenalkan oleh DeepSeek dan digunakan dalam DeepSeek-R1, dibina berdasarkan menormalkan ganjaran dalam kumpulan.
Dengan menggunakan purata kumpulan sebagai garis dasar, GRPO menggugurkan pengkritik yang dipelajari, menjimatkan memori dan mengira.
Menolak min kumpulan menjadikan respons di bawah purata mempunyai kelebihan negatif, menolak dasar daripadanya.
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.
Teruskan belajar
Lebih banyak panduan dipilih untuk topik ini
SeterusnyaPanduan seterusnya
Normalisasi Panjang dalam Pengoptimuman Keutamaan
Asas