Research

Papers and posts from the Recurvia team

  • Paper

    Paramorph Technical Report

    We introduce Paramorph, a foundation model for training control. It watches a neural network as it trains and, every hundred batches, resets the learning rate, Adam β1 and β2, dropout and weight decay according to what the run is actually doing. It is trained with MARL and is used zero-shot: a new training run gets an adaptive, run-specific schedule without a hyperparameter search.

  • Paper

    CODA - Coordination via On-Policy Diffusion for Multi-Agent Offline Reinforcement Learning

    Marcel Hedman, Kale-ab Abebe Tessera, Juan Claude Formanek et al.

    We introduce CODA (Coordination via On-Policy Diffusion for Multi-Agent Reinforcement Learning), a diffusion-based multi-agent trajectory generator for data augmentation that samples conditioned on the current joint policy, producing synthetic experience which better reflects the evolving behaviours of the agents, thereby providing a mechanism for co-adaptation.

  • Paper

    ALOHA: Agentic LLM-driven Online Hyperparameter Adaption For Reinforcement Learning

    We introduce ALOHA (Agentic LLM-driven Online Hyperparameter Adaptation), a method for online hyperparameter optimization within a single training run, that embeds a large language model directly into the RL training loop to adaptively adjust hyperparameters at fixed intervention points based on online training metrics, requiring no preprocessing and no parallel agents.