Open and completed research requests

ARC-AGI-3 tests agents on novel turn-based games whose rules, goals and controls are never stated, so an agent has to explore, work out how the environment behaves, and then act efficiently. This problem compares two agent families under one matched budget: agents that learn an explicit model of the game's dynamics and plan with it, and LLM agents that choose actions directly from observations with no explicit model. The question is which family completes more levels with fewer actions, and where each one fails.

Evaluate and improve the submitted work on graph-based reinforcement learning policies under variable action-space dimensionality, with particular focus on empirically testing whether PPO clipping behavior scales with action dimension across MuJoCo-style morphology benchmarks.

Test whether unconditional basic income schemes causally increase real GDP per capita and labor productivity in the jurisdiction that funds them, separating direct income effects from labor-supply, entrepreneurship, health, demand, and tax-financing channels.