Even if God is not limited by the capacity of His substrate, He is limited by the bandwidth of talking to ours. But I suspect that we are actually His primary substrate
RESEARCH
-
LLMs Transform Middle School Math Teaching for Diverse Learners
By
–
A recent @Stanford research highlights the potential of LLMs as vital tools for middle school math teachers to enhance the learning experience for students across diverse skill levels.
-
Neural Operators for Learning PDEs Survey
By
–
We proposed neural operators for learning PDEs. You can see a survey here https://
rdcu.be/dD6LO -

Agent Q: Multi-Agent System for Quantum Chemistry with LangGraph
By
–
The Q Agent A revolutionary LangGraph-powered multi-agent system that makes quantum chemistry accessible through natural language, achieving 87% success rate in automating complex workflows. Dive into the paper to see the future of quantum chemistry automation
-
Software as Irreducible Invariance: Spirit in Physical and Biological Systems
By
–
The pre-scientific term for software that is self organizing and self perpetuating (autopoietic) and thereby shapes and animates physics into biology is ‘spirit’. It exists (= is implemented) on many levels. Software is an irreducible invariance in reality
-
Software as Causal Pattern in Biological Systems
By
–
I‘ve come around to seeing software as the core causal pattern in biology. The software of biological systems not only controls its substrate but also homogenizes it sufficiently to allow the application of the same causal patterns over large regions of the physical domain
-
Policy Gradient for Diffusion Model Fine-tuning
By
–
It turns out we can easily apply the material we’ve covered to get the policy gradient for fine tuning a diffusion model, see eg https://
arxiv.org/pdf/2305.13301 It becomes a multi step RL problem with the reward only happening at the end. It’s not very efficient I think, but I’d love to -

Flow-GRPO: Online RL Integration into Flow Matching Models
By
–
Flow-GRPO: Training Flow Matching Models via Online RL This work introduces Flow-GRPO, the first method to embed online reinforcement learning into flow matching diffusion models. Two innovations drive it: 1. ODE → SDE conversion for richer, exploration-ready
-

ESS Measures Deep Learning Model Memory Efficiency
By
–
The performance of a Deep Learning model does not just depend on its size but on how well it makes use of its memory capacity (which in turn relates to its compressibility, destillability etc.). With ESS, LiquidAI finds a theoretically grounded measure for this