agentic-learning

Agent-G^2 introduces a Gaussian guidance framework for hint depth in reinforcement learning, enhancing performance on long-horizon agentic tasks without extra probing rollouts, with superior results on ALFWorld and WebShop benchmarks.

This paper proposes a framework for evaluating agentic learning harnesses in cybersecurity without labeled benchmarks, using a teacher-student model based on the scaling hypothesis to proxy performance improvements.

SERL-SQL proposes a selective execution-grounded reinforcement learning framework for multi-turn Text-to-SQL agents, using teacher-student likelihood gaps to reweight GRPO advantages on SQL action tokens. It achieves strong results on BIRD and Spider benchmarks.

This paper investigates whether LLM agents can infer hidden world models through interaction, finding that they struggle to build stable internal models as complexity increases.