Khushi Sharma | Searcharxiv

arXiv2026

The Stellar Mass Function of Gas-Rich Galaxies and the Underlying $M_{\rm HI}-M_{\rm star}$ Scaling Relation in the Local Universe

We estimate the Galaxy Stellar Mass Function (GSMF) of HI gas-rich galaxies using the 100\% ALFALFA ($α$) catalog, $\sim 98\%$ of which have optical counterparts in the Sloan Digital Sky Survey (SDSS) and a subset of them have counterparts in GALEX SDSS WISE Legacy Catalogue-2(GSWLC-2). We use the mass estimates from this subset which combines UV, optical and IR bands with individual dust corrections to recalibrate optical stellar mass estimates. We use a non-parametric method to estimate the GSMF of these gas-rich galaxies. The resulting, HI-selected GSMF is consistent with a single Schechter function with best-fit parameters $\left\{ϕ_* (10^{-3}\, h_{70}^{3}\,\mathrm{Mpc}^{-3}\,\mathrm{dex}^{-1}), \log_{10} (M_*/M_{\odot}) + 2\log_{10} h_{70}, α\right\} = \left\{2.30^{+0.12}_{-0.12}, \,10.83^{+0.01}_{-0.01},\, -1.14^{+0.02}_{-0.02}\right\} $. Additionally, the red and blue populations are each well described by a single Schechter function. After correcting for selection effects, we find that the red population accounts for only $\sim18\%$ of gas-rich galaxies by number, yet contributes $\sim54\%$ of the total stellar mass, with the blue population accounting for the rest. Using an optically selected sample and a joint optical-HI sample, we find gas-rich galaxies represent $\sim 33\%$ of the total stellar mass density and $\sim 39\%$ of the total galaxy number counts in the local Universe. We use the GSMF and the HI mass function (HIMF) of the HI-selected sample to obtain the $M_{\rm HI}-M_{\rm star}$ relation, which is free from selection bias.

astro-ph.GA↗

arXiv2025

Rethinking the Outlier Distribution in Large Language Models: An In-depth Study

Investigating outliers in large language models (LLMs) is crucial due to their significant impact on various aspects of LLM performance, including quantization and compression. Outliers often cause considerable quantization errors, leading to degraded model performance. Identifying and addressing these outliers can enhance the accuracy and efficiency of the quantization process, enabling smoother deployment on edge devices or specialized hardware. Recent studies have identified two common types of outliers in LLMs: massive activations and channel-wise outliers. While numerous quantization algorithms have been proposed to mitigate their effects and maintain satisfactory accuracy, few have thoroughly explored the root causes of these outliers in depth. In this paper, we conduct a comprehensive investigation into the formation mechanisms of these outliers and propose potential strategies to mitigate their occurrence. Ultimately, we introduce some efficient approaches to eliminate most massive activations and channel-wise outliers with minimal impact on accuracy.

cs.CL↗