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

2026-06-10 · Kushal Raj Bhandari, Ling Yue, Ching-Yun Ko, Dhaval Patel, Shaowu Pan, Pin-Yu Chen, Jianxi Gao arxiv

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.

📄 PDF Abstract BibTeX arXiv:2606.12674

Code (0)

등록된 구현이 없습니다.

Beyond Static Tools: Test-Time Tool Evolution for Scientific Reasoning

2026-01-12 · Jiaxuan Lu, Ziyu Kong, Yemin Wang, Rong Fu 외 arxiv

The central challenge of AI for Science is not reasoning alone, but the ability to create computational methods in an open-ended scientific world. Existing LLM-based agents rely on static, pre-defined tool libraries, a p…

Domain Adaptation

TTHE: Test-Time Harness Evolution

2026-07-09 · Jun Nie, Yonggang Zhang, Jun Song, Qianshu Cai 외 arxiv

The behavior of an LLM agent is determined not only by the underlying model, but also by its harness: the executable program that constructs context, invokes tools, verifies intermediate results, and recovers from failur…

Test-time Adaptation

CoCoDA: Co-evolving Compositional DAG for Tool-Augmented Agents

2026-05-08 · Ziyang Yu, Qiyue Li, Liang Zhao arxiv

Tool-augmented language models can extend small language models with external executable skills, but scaling the tool library creates a coupled challenge: the library must evolve with the planner as new reusable subrouti…

Mathematical Reasoning

Dynamo: Dynamic Skill-Tool Evolution for Vision-Language Agents

2026-06-29 · Yutao Sun, Yanting Miao, Hao-Xuan Ma, Mengyu Zhou 외 arxiv

Improving vision-language models (VLMs) on visual reasoning typically requires retraining or hand-designed prompts and tools. We present Dynamo, a training-free framework that adapts a frozen VLM without any weight updat…

Visual Reasoning

VehicleMemBench: An Executable Benchmark for Multi-User Long-Term Memory in In-Vehicle Agents

2026-03-25 · Yuhao Chen, Yi Xu, Xinyun Ding, Xiang Fang 외 arxiv

With the growing demand for intelligent in-vehicle experiences, vehicle-based agents are evolving from simple assistants to long-term companions. This evolution requires agents to continuously model multi-user preference…