TimePre Paper Lands in TMLR: Reversible Normalization Fixes MCL Instability in Forecasting
_vztu · x · 2026-09-23
The TimePre paper on probabilistic time-series forecasting has been accepted to TMLR. It studies combining lightweight forecasting backbones with Multiple Choice Learning (MCL), identifying scale imbalance under WTA training as the key cause of optimization instability and hypothesis collapse. A simple reversible normalization method stabilizes multi-hypothesis learning, achieving the best Distortion across all six benchmarks while keeping inference cheap via single-pass prediction. Paper and code are open-sourced.
More from Research
- MatBrain splits reasoning from tool use: two models screen 30,000 crystal candidates in 48 hours — bravo_abad · 2026-09-23
- Scale AI launches SWE-Bench Pro V2, a harder agentic coding benchmark — bigblueboo · 2026-09-23
- If AI Writes All the Papers, Peer Review Becomes Humanity's Remaining Role — sudoraohacker · 2026-09-23
- Yarin Gal: I Ignore Papers Where the Candidate Isn't First or Last Author — yaringal · 2026-09-23
- New paper: Transferring the Intelligence of VLMs to Robotic Control — _akhaliq · 2026-09-23
- NTU UMM study: generation training boosts understanding in native multimodal models, but naive sharing conflicts — jiqizhixin · 2026-09-23