AMXFP4 is an asymmetric 4-bit microscaling format that improves calibration-free LLM inference accuracy over MXFP4 and rotation-based methods.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AMXFP4: Taming Activation Outliers with Asymmetric Microscaling Floating-Point for 4-bit LLM Inference
AMXFP4 is an asymmetric 4-bit microscaling format that improves calibration-free LLM inference accuracy over MXFP4 and rotation-based methods.