Yandex Research: quantile loss may deserve to be the default for tabular regression
RichmanRonald · x · 2026-10-22
Yandex Research released a new study on tabular regression that compares 18 regression objectives and examines the role of noise and global early stopping in tabular deep learning.
A practical takeaway: quantile loss is often worth trying on regression tasks — and may even deserve to be the default. Two more announcements from the tabular team are coming soon.
Related event: Yandex Research Compares 18 Regression Objectives for Tabular Deep Learning(2 posts)→
More from Research
- EMNLP paper: diffusion LMs internally encode latent denoising timestep representations — gsarti_ · 2026-10-23
- KU Leuven and VIB open two Group Leader professorships in AI for Biology, start 2027 — jpillowtime · 2026-10-23
- Subspace Tuning curbs LLM catastrophic forgetting by freezing principal singular directions — scottniekum · 2026-10-23
- Perceptic launches PercepticOS, an AI workspace for 'scientific superintelligence' — nathanbenaich · 2026-10-23
- EPFL's Bosselut to talk long context as test-time learning at ELLIS seminar — andre_t_martins · 2026-10-23
- DRATS: Adaptive Task Sampling Boosts Worst-Task Performance in Multi-Task RL — Cohere · 2026-10-23