{"id":"arxiv/2601.03385","url":"https://arxiv.org/abs/2601.03385","tags":["cs.LG","math.PR"],"score":0,"title":"SIGMA: Scalable Spectral Insights for LLM Model Collapse","domain":"cs.LG","source":"arxiv","authors":["Yi Gu","Lingyou Pang","Xiangkun Ye","Tianyu Wang","Jianyu Lin","Carey E. Priebe","Alexander Aue"],"reviews":0,"subject":"cs.LG","summary":"The rapid adoption of synthetic data for training Large Language Models (LLMs) has introduced the technical challenge of \"model collapse\"-a degenerative process where recursive training on model-generated content leads to a contraction of distributional variance and representational quality. While the phenomenology of collapse is increasingly evident, rigorous methods to quantify and predict its onset in high-dimensional spaces remain elusive. In this paper, we introduce SIGMA (Spectral Inequalities for Gram Matrix Analysis), a unified framework that benchmarks model collapse through the…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-01-06","reactions":{}}