@berryxia: Folks, Google's latest paper has completely flipped the fundamental logic of time series forecasting. All previous models were stuck on historical data: whatever the curve does, predict that. Nexus says: prediction needs not just history, but "event context". The real reasons behind the numbers—policies, sudden events, macro trends, local shocks...

X AI KOLs Timeline Papers

Summary

Google's new paper Nexus proposes transforming time series forecasting from statistical extrapolation to multi-agent reasoning, improving prediction accuracy via event context, achieving an 86.6% reduction in MAPE on the Zillow dataset.

Folks, Google's latest paper has completely flipped the fundamental logic of time series forecasting. All previous models were stuck on historical data: whatever the curve does, predict that. Nexus says: prediction needs not just history, but "event context". The real reasons behind the numbers—policies, sudden events, macro trends, local shocks—must be interpreted together with the numbers. They neatly decomposed this using a multi-agent framework: One agent extracts event timelines from massive text, one reads macro policy, one tracks local shocks, and a final synthesizer calibrates all information with historical errors to produce the final prediction. In real-world tests, the Claude-powered Nexus version slashed the mean MAPE by 86.6% on the Zillow dataset. Not a small improvement—a devastating blow. Previously models could only "read patterns"; now they begin to "understand causality". The truly impressive part of this paper isn't a single number; it's that it completely transforms forecasting from "statistical extrapolation" into "multi-agent reasoning".
Original Article
View Cached Full Text

Cached at: 05/18/26, 02:32 PM

Guys, Google’s latest paper has completely flipped the underlying logic of time series forecasting.

Every model before was stuck on historical data: whatever the curve does, that’s what it predicts.

But Nexus says: Forecasting needs more than just history—it needs “event context.”

The real reasons behind the numbers—policies, sudden events, macro trends, local shocks—must mutually explain each other with the numbers.

They broke this down neatly using a multi-agent framework:

One agent extracts event timelines from massive text data,
One reads macro regimes,
One tracks local shocks,

And finally a synthesizer calibrates all information with historical errors to produce the final forecast.

In real-world testing, the Claude-powered Nexus version slashed average MAPE by 86.6% on the Zillow dataset.

That’s not a small improvement—it’s a dimensionality reduction strike.

Previously, models could only “recognize patterns.” Now they start to “understand causality.”

The truly impressive thing about this paper isn’t a single number—it’s that forecasting has been transformed from “statistical extrapolation” into full-on “multi-agent reasoning.”

Rohan Paul (@rohanpaul_ai):
New Google paper: A forecast needs context, not just history.

Some patterns are caused by events, not time. Nexus reframes forecasting as a reasoning problem, where events and numbers have to explain each other.

Nexus argues that forecasting improves when models read the world.

Similar Articles

Nexus : An Agentic Framework for Time Series Forecasting

Hugging Face Daily Papers

Nexus introduces a multi-agent framework that decomposes time series forecasting into specialized stages, integrating numerical patterns and contextual information using LLMs, achieving state-of-the-art results on benchmarks.

@IndieDevHailey: Google quietly releases time series nuclear weapon TimesFM: Predict the future in 5 minutes! Sales forecasting, stock price trends, website traffic, energy load, cryptocurrency volatility... These headache-inducing future numbers now have a unified answer. TimesFM: → Trained on 100 billion real-world time series data...

X AI KOLs Timeline

Google has released TimesFM, a time series forecasting model trained on 100 billion real-world time series data, supporting zero-shot prediction. It is free, open-source, and can run locally on ordinary computers.