Data-Free Discovery of Critical LLM Parameters
December 2025
Post-training quantization of large language models introduces errors that degrade quality, but not all weights contribute equally to this degradation. PeakWeights identifies critical weights in a single forward pass without calibration data. Weight magnitude multiplied by peak activation closely tracks worst-case quantization error. Across five architectures, protecting 50 weights during 4-bit quantization recovers 61-99% of lost perplexity.
| Model | FP16 | 4-bit | +PeakWeights | Recovery |
|---|---|---|---|---|
| Qwen2.5-7B | 10.62 | 11.51 | 10.63 | 99% |
| SmolLM2-1.7B | 17.92 | 24.56 | 17.99 | 99% |
| Phi-3-mini | 11.65 | 12.67 | 11.68 | 97% |
| DeepSeek-R1-7B | 70.73 | 73.52 | 70.91 | 93% |
| Mistral-7B | 13.44 | 13.80 | 13.58 | 61% |
Perplexity on WikiText-103. Lower is better. Recovery = (4bit - protected) / (4bit - FP16).