DerOeko

Scalable Online RL Training System (vLLM + RLAIF)

Note: This gist outlines architectural patterns and infrastructure used for research. Source code and proprietary datasets are excluded per lab data policy.

Technical Highlights

Online RL Implementation

• GRPO (Group Relative Policy Optimization): Implemented group-based baselines for multi-turn agents, removing the need for a separate value network (Critic).
• Loss Variants: Support for DAPO, BNPO, and DR-GRPO with token-level masking for long-context reasoning.
• Curriculum Learning: Dynamic CurriculumDatasetCallback to shift training distributions from simple to complex tasks to prevent collapse.

DerOeko

Distributed Bayesian Optimization & Cognitive Modeling Framework

Note: This gist outlines architectural patterns and infrastructure used for research. Source code and proprietary datasets are excluded per lab data policy.

Technical Highlights

Reinforcement Learning Environment

• Gymnasium Wrappers: Custom observation-based environments with flexible state transitions for cognitive task modeling.
• Configurable Rewards: Decoupled reward structure from environment logic to allow rapid iteration on task definitions.