DN-MOPD
Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation. Data and models.
Viewer • Updated • 26.7k • 172Note Training prompt sets: student (2,700) and GRPO teachers (math/code/IF, 8,000 each)
Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation
Paper • 2609.35347 • Published • 176Note Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation
XINLI1997/DN-MOPD-Qwen3.5-9B
Image-Text-to-Text • 9B • Updated • 10Note DN-MOPD student, 80 updates (paper Tables 1-2)
XINLI1997/DN-MOPD-Qwen3.5-4B
Image-Text-to-Text • 5B • Updated • 12Note DN-MOPD student, 80 updates (paper Tables 1-2)
XINLI1997/DN-MOPD-Qwen3.5-2B
Image-Text-to-Text • 2B • Updated • 12Note DN-MOPD student, 80 updates (paper Tables 1-2)
XINLI1997/DN-MOPD-Qwen3.5-9B-teacher-math
Image-Text-to-Text • 9B • Updated • 11Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-9B-teacher-code
Image-Text-to-Text • 9B • Updated • 8Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-9B-teacher-if
Image-Text-to-Text • 9B • Updated • 10Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-4B-teacher-math
Image-Text-to-Text • 5B • Updated • 13Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-4B-teacher-code
Image-Text-to-Text • 5B • Updated • 8Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-4B-teacher-if
Image-Text-to-Text • 5B • Updated • 14Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-2B-teacher-math
Image-Text-to-Text • 2B • Updated • 13Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-2B-teacher-code
Image-Text-to-Text • 2B • Updated • 14Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-2B-teacher-if
Image-Text-to-Text • 2B • Updated • 11Note GRPO expert used as a teacher
XINLI1997/DN-MOPD-Qwen3.5-9B-baseline-label
Image-Text-to-Text • 9B • Updated • 12Note Baseline: MOPD with label routing, 80 updates
XINLI1997/DN-MOPD-Qwen3.5-4B-baseline-label
Image-Text-to-Text • 5B • Updated • 17Note Baseline: MOPD with label routing, 80 updates
XINLI1997/DN-MOPD-Qwen3.5-2B-baseline-label
Image-Text-to-Text • 2B • Updated • 14Note Baseline: MOPD with label routing, 80 updates
XINLI1997/DN-MOPD-Qwen3.5-9B-160updates
Image-Text-to-Text • 9B • Updated • 15Note DN-MOPD student, 160 updates (paper Table 5)
XINLI1997/DN-MOPD-Qwen3.5-4B-160updates
Image-Text-to-Text • 5B • Updated • 15Note DN-MOPD student, 160 updates (paper Table 5)
XINLI1997/DN-MOPD-Qwen3.5-2B-160updates
Image-Text-to-Text • 2B • Updated • 14Note DN-MOPD student, 160 updates (paper Table 5)
XINLI1997/DN-MOPD-Qwen3.5-9B-baseline-label-160updates
Image-Text-to-Text • 9B • Updated • 17Note Baseline: MOPD with label routing, 160 updates (paper Table 5)
XINLI1997/DN-MOPD-Qwen3.5-4B-baseline-label-160updates
Image-Text-to-Text • 5B • Updated • 16Note Baseline: MOPD with label routing, 160 updates (paper Table 5)
XINLI1997/DN-MOPD-Qwen3.5-2B-baseline-label-160updates
Image-Text-to-Text • 2B • Updated • 16Note Baseline: MOPD with label routing, 160 updates (paper Table 5)