TW3Cast: A Frozen Router of Lightly Fine-Tuned Foundation Models for Time-Series Forecasting on GIFT-Eval, Selected Entirely on the Training Split
Source
Nathan Thierry, Andre-Louis Rochet
Author
Nathan Thierry, Andre-Louis Rochet
Date
Key takeaways · AI-distilled
TW3Cast ranked 3rd of 130 entries on GIFT-Eval by mean MASE rank as of 2026-09-14. The two entries above it use agents or language models; TW3Cast uses neither.
For each of 97 dataset, frequency and horizon configurations, a frozen table serves one of four modes: a fine-tuned Chronos-2, TiRex or Toto specialist, a quantile blend containing a specialist, a blend of base models, or a backtest tournament.
Every choice is made on a backtest carved from the training split. guardrailsThe checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.Full definition → include a joint accuracy and calibrationHow well a model's confidence matches reality — a calibrated model saying "90% sure" is right about 90% of the time.Full definition → criterion, an asymmetric margin against candidates that saw the series in training, and conservative per-window gates.
The best single base model reaches a mean MASE rank of 33.8 and the tournament alone 38.0, against 19.4 for the full router. The authors say every leaderboard number regenerates from released artifacts with one script.
Terms in this piece · Glossary
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
guardrails — The checks around a model that block bad inputs and outputs — filters, validators, and permission rules the model itself can't override.
calibration — How well a model's confidence matches reality — a calibrated model saying "90% sure" is right about 90% of the time.
Why it matters
Shows that for time-series forecasting, a frozen ensemble can beat most LLMA large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.Full definition →-driven agentic systems, a useful data point before adding AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → orchestration overhead to a forecasting pipeline.
articleWhen Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability RoutingYufeng Wang
articleAutoFyn Technical Report: Non-Parametric Expert Iteration for Long-Horizon AgentsAdib Hasan, Daniel Schaffield, Akashnil Dutta, Tarik Adnan Moon