TW3Cast: A Frozen Router of Lightly Fine-Tuned Foundation Models for Time-Series Forecasting on GIFT-Eval, Selected Entirely on the Training Split
- What changed
- TW3Cast achieves rank 3 on GIFT-Eval using a frozen router of fine-tuned models without agents.
- Why you should care
- Frozen routing of fine-tuned models achieves rank 3 without LLMs.
- Your move
- Watch. Non-agent routing outperforms single models on GIFT-Eval.
- Evidence
- Confirmedarxiv.org ↗
- What to watch next
- Independent replication of TW3Cast results on other time-series benchmarks.
- Event
- research
- Event date
- Sep 25, 2026
- Relevant to
- General AI readers