Evoflux: Inference-Time Evolution of Executable Tool Workflows for Compact Agents

Researchers introduced Evoflux, an inference-time evolutionary search method that repairs executable tool workflows for compact language model agents. On MCP-Bench tasks involving 250 live tools, Evoflux improved execution feasibility from roughly 3% to 17-24% across small planners, outperforming supervised fine-tuning and DPO approaches. The method demonstrates that execution-grounded search is more reliable than teacher-trace distillation for enabling small agents to handle tool discovery, parameter validation, and dependency tracking.

arXiv:2606.12674v1 Announce Type: new Abstract: Compact language models LMs reduce cost, latency, and deployment risk for tool agents. Yet MCP-style tool use requires more than isolated function calling: an agent must discover tools from live catalogs, satisfy schemas, preserve dependencies across intermediate outputs, and ground final responses in executed evidence. Small planners often generate plausible workflow graphs that fail under tool resolution, parameter validation, dependency tracking, or execution. We argue that this failure mode is poorly handled by small-corpus distillation. A few hundred teacher traces can teach workflow format, but rarely cover the recovery behavior needed to repair failed plans over changing tool catalogs. We introduce Evoflux, an inference-time evolutionary search method that treats compact tool use as the repair of executable tool workflows. It evolves typed workflow graphs through structured edits, execution feedback, adaptive intensity, meta-guided redesign, and diversity pruning. On held-out MCP-Bench tasks spanning live MCP servers and 250 tools, Evoflux raises execution feasibility from roughly 3% to 17-24% across small planners. In contrast, SFT and SFT+DPO on the same search-mined data match, underperform, or collapse below zero-shot performance; ReAct reaches higher peaks, but with higher variance and token cost. These results show that execution-grounded search is more reliable under scarce teacher-trace budgets.