Search

Any time Past 24 hours Past week Past monthSort:Relevance Newest

Benchmark

WikiText-2

Reported in 19 papers on Zotpaper. The latest, each as the paper states it:

Combining model compression techniques can collapse small language models920.56 perplexity
4-bit language models run as integer math on Apple's M5 chips without extra accuracy loss10.4 perplexity

WikiText2 (2)All benchmarks →

2 results for “WikiText2”

Research · Efficiency & systems
Language models can be trained for compression nearly 13 times faster
SQuAT: Spherical Lattice Vector Quantization-Aware Training
Loeschcke et al. decompose weight blocks into direction and magnitude and apply factored straight-through estimators, improving accuracy over joint-STE baselines and matching scalar INT-QAT below 2.5 bits per weight.
WikiText2: 14.76 perplexity (as reported)
23 Oct 2026University of Copenhagen, Qualcomm AI Research
Research · Efficiency & systems
Two-bit scalar quantization better preserves language model predictions
FidelityQ: Preserving Predictive Fidelity in Ultra-Low-Bit Language Model Quantization
FidelityQ combines adaptive four-level weight coding with targeted teacher-guided calibration to improve accuracy while retaining simple integer-compatible decoding.
WikiText2: 6.96 perplexity (as reported)
15 Oct 2026The Hong Kong Polytechnic University, Huawei Technologies Ltd., The University of Hong Kong