ReactiveBayes

🎉 Big news: RxInfer.jl is now a @NumFOCUS Affiliated Project! Born from PhD research on reactive message passing, it’s now an open-source tool for fast, scalable Bayesian inference — and part of a world-class ecosystem #JuliaLang #BayesianInference #NumFOCUS #OpenSource

This is a cool Julia version of my Jax library for Bayesian Structural Time Series modeling (github.com/probml/sts-jax) from the folks at @ReactiveBayes. It can easily handle non-linear and non-conjugate likelihoods (eg Poisson distribution for integer count observations). For inference, it uses a very fast variational message passing algorithm (combined with data augmentation and local MC approximations for non-conjugate messages). In this example, they infer the latent states of the SSM as well as the model parameters (representing how the continuous latent demand evolves over time). But their RxInfer.jl library also supports inference in many other kinds of graphical model (see examples.rxinfer.com).

New Bayesian Structured Time Series example: predicting taco demand during #NeurIPS 2025 ⚡️ Learnable Dynamics 🔢 Non-Conjugacies 🏎️ Blazing Speed examples.rxinfer.com/categories/adv… #JuliaLang #BayesianInference #DataScience

@sirbayes This feels RxInfer.jl-ready: a custom node + a few bespoke update rules! Congrats on the NeurIPS paper!

I am pleased to share our new NeurIPS paper for online Bayesian inference for neural networks. Instead of focusing on updating the parameter posterior, we work with the predictive posterior (which makes much more sense for non-identifiable models, and gives us more algorithmic freedom for developing faster methods).

🏆 Best Technical Paper at @iwai_ws! Recasting EFE planning as Variational Inference → sample-efficient Active Inference agents that actually run. Built in RxInfer. Won by @wouterwln. Paper: arxiv.org/abs/2508.02197 Code: github.com/biaslab/EFEasV… #IWAI2025 #RxInfer #Bayesian

🚀 We’re open-sourcing Gears.jl — a Julia package for precise task scheduling in simulation environments! ⚙️ Run timed, event-based, or ASAP jobs with flexible schedulers, multiple clocks, and multi-threading. 🔗 github.com/lazydynamics/G… #JuliaLang #OpenSource #Agents

@sonoftheright Apologies and thanks! discord.gg/TtdStfnjNn

@ReactiveBayes Invite invalid! Looks like it expired...

We’re on Discord! You asked for better community engagement around RxInfer & RxInfer Pro—so here it is. Join us for Q&A, updates, hackathons & more. It’s quiet now, but let’s build it together. 📷 discord.gg/UvMhrbQh #RxInfer #BayesianInference #ProbabilisticProgramming

We've been seeing some cool examples of fast Bayesian inference + LLMs - from trust/reliability to LLMs participating in inference (even learning to derive variational rules!). Check out the examples: examples.rxinfer.com

We’re on Discord! 🎧🚀 Moving beyond GitHub discussions—join the new RxInfer server (by @LazyDynamics ) to chat, ask Qs, get updates, and plan hackathons. Jump in: discord.gg/UvMhrbQh #RxInfer #BayesianInference #ProbabilisticProgramming

In Active Inference, a lot of time is spent on computing Expected Free Energy. What if we could tweak the generative model such that EFE can be minimised with traditional variational inference methods?

🧠 We hooked LLMs into Bayesian inference in 30 minutes ChatGPT as a native node in message-passing. Working sentiment clustering with uncertainty quantification. 📖 Notebook: examples.rxinfer.com/categories/exp… #BayesianInference #LLM

@whabs99 @NumFOCUS FFGs are hypergraphs by definition - they already handle complex multi-variable dependencies through factor nodes. See Wainwright & Jordan pg250 for the mathematical foundations.

@ReactiveBayes @NumFOCUS Any thoughts on using hyper edges between nodes and clusters for complex dependency active inference for graph based networks ?

🎉 Big news: RxInfer.jl is now a @NumFOCUS Affiliated Project! Born from PhD research on reactive message passing, it’s now an open-source tool for fast, scalable Bayesian inference — and part of a world-class ecosystem #JuliaLang #BayesianInference #NumFOCUS #OpenSource

@fchollet There's a great toolbox designed for Active Inference 😉

The Free Energy Principle is a pretty good idea, but its core value isn't to serve as a grand unifying "theory of everything" for cognition. Rather, its core insight is the rigorous emphasis on active inference, which has been badly missing from the deep learning era.

rxinfer.com can be powerful for robotics - but requires deep probabilistic programming knowledge. What if there was a simpler way? LLM → RxInfer → Robotics = 𝗖𝘂𝗿𝘀𝗼𝗿 𝗳𝗼𝗿 𝗿𝗼𝗯𝗼𝘁𝗶𝗰𝘀? #RxInfer #Robotics #LLM #CursorForRobotics #AIRobotics

Check out our DeepWiki page - it looks pretty accurate! deepwiki.com/ReactiveBayes/… Thanks to @cognition_labs!

@LazyDynamics @bayesgroup So cool!

Great visit with the @bayesgroup in Bremen! 🚀 So many inspiring discussions on dynamic systems & probabilistic methods. Huge thanks to Dmitry Vetrov & the whole team for the warm welcome. Excited for future collaborations! #Collaboration #Innovation #BayesianML

🎯 New Contextual Multi-Armed Bandits example in RxInferExamples.jl! Fresh take using message passing inference. No MCMC, no hyperparameters - just principled decision making. examples.rxinfer.com/categories/bas… #BayesianML #JuliaLang #RxInfer

100%.

🚀Behind the scenes at Lazy Dynamics: We are working on Cortex, our future reactive message-passing backend for RxInfer! Designed for fast, reliable, and asynchronous probabilistic inference on edge devices. Ideal for robotics, audio processing, and more. github.com/ReactiveBayes/…