Twitxy Saved

Chris(t)ensor

Verified account This account is verified. Learn more

@tensorlink_dev

Building on bittensor - sn91 (@cascade_sn91)
Planum Temporale Joined January 2025
250 Following 249 Followers 171 Posts

Latest posts 5

Chris(t)ensor Verified account @tensorlink_dev
@numinous_ai is producing data that will give true edge in prediction markets, trading and beyond. This is a positive feedback loop - we help build better graphs they help train better TSFMs
SN91, Cascade Verified account @cascade_sn91 5d
Track how the world changes. Predict how it evolves.

We’ve partnered with @numinous_ai to combine knowledge graphs with time-series forecasting.

Numinous maps relationships between companies and events, and predicts how those relationships may change. Cascade brings forecasting models to assist in flagging future activity, pointing agents towards worthy developments.

In turn, Numinous’s graph will help us develop synthetic training data for our mixture-of-experts model, capturing how changes at one company can ripple through connected businesses.

Together, we’re working towards better forecasts of both changing relationships and the numerical trends they influence.
View quoted post on X
Chris(t)ensor Verified account @tensorlink_dev
Extremely powerful approach to modelling, forecasting and knowledge distillation
Numinous Verified account @numinous_ai Oct 22
Today starts our open-ended forecasting competition where miners forecast the future state of a dynamic knowledge graph. First focused on the top US equities and their supply chain. Tomorrow it will also have a geopolitics graph.
View quoted post on X
Chris(t)ensor Verified account @tensorlink_dev
Road to SOTA
SN91, Cascade Verified account @cascade_sn91 Oct 22
From 4M to 1B+. Cascade’s biggest upgrade yet.

We’re preparing to train Cascade-MoE: a mixture-of-experts time-series foundation model targeting 1B+ total parameters.

We’ve shown that synthetic data generated by Bittensor miners can train a 4M-parameter forecasting model from scratch, reaching parity with Toto 2.0 4M. Now we’re scaling up, with SOTA forecasting as our goal.

Cascade-MoE will use a learned router to select combinations of expert networks based on patterns in the input, alongside an always-active shared expert. This gives the model access to specialised capacity without activating every stored weight for every input.

We’re also evolving the competition so miners can generate synthetic data targeted at particular domains and time resolutions, such as hourly versus daily series.

The architecture is designed to support targeted training of selected expert subsets, while the subnet tests whether contributions improve forecasting performance.

The goal: SOTA forecasting accuracy, built through competition on #SN91.
View quoted post on X