Someone finally indexed the time-series forecasting flood

What it does

This repository is a manually curated index of research papers, benchmarks, datasets, courses, and code repositories focused on time series forecasting and deep learning. The maintainer organizes papers by year—from 2017 through 2026—and links directly to arXiv preprints and official implementations when available. It also collects practical resources like financial benchmarks (FinTSB), evaluation leaderboards (GIFT-Eval), and tutorials. Think of it as an awesome-list for a field that publishes a new architecture every Tuesday.

The interesting bit

The value is in the filtering: instead of a flat dump, the list separates foundational benchmarks from the annual paper deluge and explicitly tags whether official code exists. That saves you from the classic “promising abstract, missing repo” trap.

Key highlights

• Chronological paper index from 2017 to 2026, with direct links to arXiv and official code.
• Curated benchmark section including FinTSB, GIFT-Eval, QuitoBench, and TIME.
• Additional sections for datasets, courses, tutorials, blogs, and books.
• Explicitly lists applications like Nixtla’s TimeGPT alongside academic research.
• Updated regularly; commit activity and closed-issue badges suggest active maintenance.

Caveats

• The repository contains no original code or libraries; it is purely a reading list and index.
• The README runs very long; the provided source is truncated, so sections beyond the paper list are only partially visible.
• Several top entries carry 2026 publication dates, suggesting the list includes forthcoming preprints.

Verdict

Worth bookmarking if you research or engineer forecasting models and need a quick scan of the latest architectures and benchmarks. Skip it if you are looking for a drop-in Python library or reusable training pipeline.