Entity · paper

The Illusion of Equivalency: Statistical Characterization of Quantization Effects in LLMs

paperactiveprovisionalthe-illusion-of-equivalency-statistical-characterization-of-quantization-effects-in-llms-d237e69a·1 events·first seen 6h ago

Aliases: The Illusion of Equivalency: Statistical Characterization of Quantization Effects in LLMs

Co-occurring entities

correctness agreement

More like this (12)

Lloyd-Max quantization Janus: A Benchmark for Goal-Conditioned Information Distortion in LLMs quantization-induced degradation How Much is Left? LLMs Linearly Encode Their Remaining Output Length Continual LLM Upcycling: A Predictor-Gated Bank-Wise Sparsity Training Recipe for Dense-to-Sparse LLMs On The Effectiveness-Fluency Trade-Off In LLM Conditioning: A Systematic Study Operadic consistency: a label-free signal for compositional reasoning failures in LLMs code synthesis LLMs Knowledge Knows, Verbalization Tells: Disentangling Latent Directions for Mathematical Solvability in LLMs Can LLMs Judge Better Than They Generate? Evaluating Task Asymmetry, Mechanistic Interpretability and Transferability for In-Context QA Language Models Compare Quantities Using Number-specific and Unit-specific Heuristics Quantum vs. Classical Machine Learning: A Unified Empirical Comparison

Recent events (1)

6arXiv · cs.AI·6h ago·source ↗

New paper exposes behavioral divergence in quantized LLMs beyond accuracy/perplexity metrics

A new arXiv preprint argues that standard accuracy and perplexity metrics fail to capture behavioral changes induced by post-training quantization. The authors introduce 'correctness agreement', a decision-level metric measuring overlap in correct predictions between base and quantized models, and find behavioral divergence emerges even when task performance appears preserved. Analysis of attention weight distortions reveals non-linear breakpoints at low bit-widths and that query/key projections are more sensitive to quantization than value/output projections. The findings challenge the assumption that quantized models are behaviorally equivalent to their base counterparts.

Evaluation and Benchmarking Inference Economics correctness agreement The Illusion of Equivalency: Statistical Characterization of Quantization Effects in LLMs