Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Training Iteration Efficiency on Qwen3-0.6B base (train)

100.2Per-step Latency (ms)

FORGE

81.016210.508340469.492Jun 22, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.06
100.24.9
2026.06
107.24.9
2026.06
107.36.7
2026.06
111.97.3
2026.06
1186.7
2026.06
125.37.3
2026.06
134.64.2
2026.06
135.24.7
2026.06
136.64
2026.06
1534
2026.06
159.14.2
2026.06
162.24.7
2026.06
168.34.2
2026.06
177.46.3
2026.06
179.84.2
2026.06
195.16.3
2026.06
527.13.1
2026.06
579.83.1