AI RESEARCH
Does RL Expand the Capability Boundary of LLM Agents? A PASS@(k,T) Analysis
arXiv CS.LG
•
ArXi:2604.14877v1 Announce Type: new Does reinforcement learning genuinely expand what LLM agents can do, or merely make them reliable? For static reasoning, recent work answers the second: base and RL pass curves converge at large k. We ask whether this holds for agentic tool use, where T rounds of interaction enable compositional strategies that re-sampling cannot recover. We