academy.dair.ai

Google's LEAP framework wraps a general-purpose LLM in an agentic scaffold that grounds every step in the Lean compiler and iterates against verifier feedback. It solves all 12 Putnam 2025 problems with one model, lifting the one-shot solve rate of the Lean-IMO-Bench from under 10% to 70%, outperforming a specialized gold-medal system that scores 48. Paper: arXiv:2606.03303. Learn to build effective AI agents at academy.dair.ai.

Meta proposes AIRA, a dual-agent system that autonomously discovers neural architectures outperforming Llama 3.2 at 350M, 1B, and 3B scales within a 24-hour compute budget, offering a reusable engineering paradigm for AI agent design.

The article argues that agentic AI systems are more promising for achieving AGI than simply scaling up foundation models.

Paper reveals significant discrepancy between model's recognition and execution of tool calls, with match rate 26-54%, concentrated in cognition-to-action transition.

This tweet promotes an AI learning platform and plugin, lacking technical depth or practical information.