Mixed-Precision Quantization
Cross-source consensus on Mixed-Precision Quantization from 1 sources and 5 claims.
1 sources · 5 claims
Uses
How it works
Comparisons
Highlighted claims
- At 2.25 average bits, RCO achieved lower perplexity than EvoPress, IMPQ, and HIGGS on the reported datasets. — Budget Constraints as Riemannian Manifolds
- At 2.5 bits, RCO had similar quality to EvoPress but much lower wall-clock time. — Budget Constraints as Riemannian Manifolds
- For Qwen3-8B quantization, RCO optimized layer-bitwidth assignment after candidate bitwidths were pre-quantized with GPTQ. — Budget Constraints as Riemannian Manifolds
- Surrogate methods were closer to RCO at higher bitwidths because compression was easier. — Budget Constraints as Riemannian Manifolds
- RCO is especially relevant for high-compression quantization where proxy objectives fail. — Budget Constraints as Riemannian Manifolds