Prox model achieves expert mixture decoding without calibration data
The story so far
The Prox model accelerates expert mixture reasoning through top-k channel selection and weighted readers while maintaining accuracy, without calibration data. Under 13% and 18% budgets, Prox's top-k version scored better than or equal to the threshold version in 40 out of 43 comparisons, including all 23 generation tasks, and avoided the repetitive loop crashes caused by calibration thresholds on Gemma 4.
AI-written from the reports · Updated 20 days ago
Coverage timeline
Follow the coverage to see every side of the story.
- Hugging Face Daily PapersChannel Selection for Mixture-of-Experts Decoding Needs No Calibration Data
The Prox model accelerates expert mixture inference through top-k channel selection and weighted reader, while maintaining accuracy without calibration data. Under 13% and 18% budgets, the top-k version of Prox outperformed or matched the threshold version in 40 out of 43 comparisons, including all 23 generation tasks, and avoided the repetitive loop crashes caused by calibration thresholds on Gemma 4.