Models trained as a part of the "Attention Drift: What Speculative Decoding Models Learn" paper, shared for reproducing experiments.
Doğaç Eldenk
Dogacel
AI & ML interests
None yet
Recent Activity
liked a model about 22 hours ago
nvidia/gpt-oss-120b-Eagle3-v3 liked a dataset about 23 hours ago
nvidia/SPEED-Bench updated a model 1 day ago
Dogacel/specdrift-gpt-oss-20b-eagle3Organizations
None yet
models 21
Dogacel/specdrift-gpt-oss-20b-eagle3
0.4B • Updated • 50 • 1
Dogacel/specdrift-gpt-oss-120b-eagle3
0.4B • Updated • 215
Dogacel/specdrift-qwen3.6-35b-a3b-eagle3
0.2B • Updated • 37 • 1
Dogacel/specdrift-qwen3.6-27b-eagle3
0.6B • Updated • 52
Dogacel/paper-attention-drift-qwen35-9b-post
0.4B • Updated • 12
Dogacel/paper-attention-drift-qwen35-9b-pre
0.4B • Updated • 35
Dogacel/paper-attention-drift-gpt-20b-pre
0.3B • Updated • 72
Dogacel/paper-attention-drift-gpt-20b-post
0.3B • Updated • 71
Dogacel/paper-attention-drift-qwen3-pre
0.4B • Updated • 26
Dogacel/paper-attention-drift-qwen3-post
0.4B • Updated • 39
datasets 4
Dogacel/nemotron-post-training-v2-gpt-oss-20b-regen
Viewer • Updated • 931k • 25
Dogacel/nemotron-post-training-v2-gpt-oss-120b-regen
Viewer • Updated • 930k • 34
Dogacel/nemotron-post-training-v2-qwen-3.5-0.8b-regen
Viewer • Updated • 932k • 38
Dogacel/nemotron-post-training-v2-qwen-3.5-9b-regen
Viewer • Updated • 940k • 27