←── back to feed
/topics/arxiv-cs-lg-papers-may-27-2026

arXiv cs.LG papers May 27 2026

15 items1 sourcesupdated 21d agotrend 0

On May 27, 2026, 15 new machine learning papers were posted to arXiv's cs.LG category, covering topics ranging from data curation and quantization for large language models to anomaly detection, federated learning, weather forecasting, and medical signal classification. The papers address practical deployment challenges including output constraints for small models, low-bit quantization efficiency, and contamination auditing in foundation models.

  • GEM reformulates LLM data curation as a variational problem on the hypersphere using Minorize-Maximize optimization to address embedding anisotropy.
  • AirCast-SR downscales global weather forecasts from 0.25 degree (~28 km) to 1 km resolution using latent consistency diffusion for 67-hour predictions.
  • InfoQuant shapes activation distributions for low-bit LLM quantization by matching distributions to uniform quantizers rather than just suppressing outliers.
  • ARBITER identifies wrong-majority failures in test-time sampling where correct answers are outvoted by clustering reasoning trajectories into basins.
  • TSFMAudit introduces the first pretraining contamination auditing framework for time series foundation models to detect evaluation dataset exposure.
  • HRVConformer uses hybrid Convolution-Transformer architecture to classify neonatal hypoxic-ischemic encephalopathy directly from raw heart rate signals.
[BLG]blog/rss15
GEM: Geometric Entropy Mixing for Optimal LLM Data Curation
arXiv cs.LG · Yue Min, Ziyun Qiao, Ruining Chen, Yujun Li · 21d
The Constraint Tax: Measuring Validity-Correctness Tradeoffs in Structured Outputs for Small Language Models
arXiv cs.LG · Jaideep Ray · 21d
AirCast-SR: A Foundation Model for Kilometer-Scale Atmospheric Super-Resolution via Latent Consistency Diffusion
arXiv cs.LG · Somnath Luitel, Manmeet Singh, Joshua Durkee, Abdullah Al Fahad, Naveen Sudharsan, Prabhjot Singh, Cenlin He, Harsh Kamath, Zong-Liang Yang, Krishnagopal Halder, Sandeep Juneja, Parthasarathi Mukhopadhyay, Saptarishi Dhanuka, Amit Kumar Srivastava · 21d
SilIF: Silhouette-Augmented Isolation Forest for Unsupervised Transaction Fraud Detection
arXiv cs.LG · Venkatakrishnan Gopalakrishnan · 21d
Neural Bayesian Sequential Routing
arXiv cs.LG · Yongchao Huang · 21d
TSFMAudit: Data Contamination Auditing in Forecasting Time Series Foundation Models
arXiv cs.LG · Hongkai Li, Shifeng Xie, Lefei Shen, Zhuo Li, Mouxiang Chen, Xiaobin Zhang, Han Fu, Jianling Sun, Xiaoxue Ren, Chenghao Liu · 21d
On the Push-Based Asynchronous Federated Learning: A Bias-Correction Aggregation Approach
arXiv cs.LG · Jiahui Bai, Hai Dong, A. K. Qin · 21d
Planning Neural Dynamics with Lie Group Embedding through Supervised Projective Manifold Learning
arXiv cs.LG · Tianwei Wang, Bryan Chen, Qian Zuo, Qiyue Xia, Xin Li, Wei Pang · 21d
When Rule Violations Are Rare: Chimera Training for Logical Anomaly Detection
arXiv cs.LG · Alejandro Ascarate, Leo Lebrat, Rodrigo Santa Cruz, Clinton Fookes, Olivier Salvado · 21d
ARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time Sampling
arXiv cs.LG · Meng Cai, Lars Kulik, Farhana Choudhury · 21d
InfoQuant: Shaping Activation Distributions for Low-Bit LLM Quantization
arXiv cs.LG · Ke Li, Dong An, Xiaoling Zang, Can Ye, Liang Xie, Qibo Qiu, Chen Shen, Xiaofei He, Wenxiao Wang · 21d
GAC: Noise-Aware Adaptive Mixing for Hybrid SFT-RL Post-Training
arXiv cs.LG · Yuelin Hu, Zhenbo Yu, Zhengxue Cheng, Wei Liu, Li Song · 21d
Max-Window Scale Estimation for Near-Lossless HiF8 W8A8 Quantization-Aware Training
arXiv cs.LG · Yingying Cheng, Jinquan Shi, Li Zhou, Zhiyang He, Zhaoyi Sun, Fan Zhang, Jie Sun · 21d
HRVConformer: Neonatal Hypoxic-Ischemic Encephalopathy Classification from the Heart Rate signals
arXiv cs.LG · Shuwen Yu, William P Marnane, Geraldine B. Boylan, Gordon Lightbody · 21d
Modeling Dynamic Mixtures of Time-Delay Systems from Streaming Time Series
arXiv cs.LG · Ren Fujiwara, Yasuko Matsubara, Yasushi Sakurai · 21d