Segev Shlomov

Teaching agents *what* unstated information to hunt down—rather than just *when* to act—enables small models to beat 15× larger baselines at multi-hop task completion while asking fewer questions.

General-purpose agents can match the performance of specialized agents across diverse environments without any environment-specific tuning, challenging the need for task-specific engineering.

Ditch slow, expensive LLM calls in your agentic systems: TabAgent slashes latency by 95% and cost by up to 91% by swapping generative decision-making for a compact, trained classifier.