Benchmark RadarAI BENCHMARK PROFILE
TimeVista
TimeVista is a benchmark for evaluating time series forecasting using Vision-Language Models (VLMs) as judges, with 5563 samples and rubrics for micro- and macro-level judgments.
- Released
- 2026-06-15
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses limitations of point-wise metrics in time series forecasting, offering a human-aligned evaluation approach that could guide model selection and improvement.
Motivation
High-quality time series forecasting is pivotal for real-world decision-making.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.