Introducing Delphi
Delphi lets you watch machine learning models compete live on benchmarks and buy a stake in those you think are best, creating the first live market signal of model performance.
Gensyn
CodeZero: Extending RL-Swarm Toward Cooperative Coding Agents
CodeZero extends Gensyn's RL-Swarm framework into the domain of code
Introducing CodeAssist
Today, we're introducing CodeAssist, an AI coding assistant that trains on your local machine. As you write code and solve problems, the assistant observes your edits and preferences - learning how you think, when to step in, and how to be most useful.
SAPO, Efficient LM Post-Training with Collective RL
This is an academic paper describing SAPO, a meta-algorithm that wraps around your preferred policy gradient algorithm.
Introducing Judge
Judge brings cryptographically verifiable AI evaluation to scale. Built on Verde, Judge ensures independent verification - eliminating opaque APIs.
Introducing BlockAssist
BlockAssist is an AI Minecraft assistant that learns from your in-game actions, enabling reinforcement learning research in an interactive environment.
Introducing RL Swarm’s new backend: GenRL
GenRL is a new framework designed from the ground up to simplify and accelerate the creation of advanced RL environments, particularly those involving multiple agents.
CheckFree: fault tolerant training without checkpoints
This is an academic paper describing CheckFree, a novel recovery method for failures in distributed training that does not require checkpointing or redundant computation.
NoLoCo: training large models with no all-reduce
This is an academic paper describing NoLoCo, a novel optimisation method for distributed training that replaces the global synchronisation step with a gossip method.
Diverse Expert Ensembles: embarrassingly parallel LLMs from diverse experts
This is an academic paper that finds benefits to heterogeneity (different model sizes and number of training steps) when training embarrassingly-parallel ensembles of expert models.
RL Swarm: a framework for collaborative RL
This is open source code (MIT Licence) for peer-to-peer nodes that perform collaborative reinforcement learning over the internet, accessible by anyone on consumer or datacentre hardware.
SkipPipe: a communication efficient method for decentralised training
This is an academic paper for efficient communication in pipeline parallel training. It introduces an optimal scheduling algorithm that maximises performance and fault tolerance whilst minimising convergence impact from layer skips.