Mock Worlds, Real Skills: Building Small Agentic Language Models with Synthetic Tasks, Simulated Environments, and Rubric-Based Rewards

TL;DR

This paper introduces SYNTHAGENT, a framework for training small language models with synthetic tasks, simulated environments, and rubric-based rewards, leading to improved agentic capabilities across diverse tasks.

Contribution

The paper presents SYNTHAGENT, a novel approach that synthesizes diverse training data and simulates environments to enhance small language models' agentic skills.

Findings

Small models outperform larger baselines on 14 datasets.

Synthetic data training improves task performance significantly.

Models actively query users for missing information.

Abstract

Small LLMs often struggle to match the agentic capabilities of large, costly models. While reinforcement learning can help, progress has been limited by two structural bottlenecks: existing open-source agentic training data are narrow in task variety and easily solved; real-world APIs lack diversity and are unstable for large-scale reinforcement learning rollout processes. We address these challenges with SYNTHAGENT, a framework that jointly synthesizes diverse tool-use training data and simulates complete environments. Specifically, a strong teacher model creates novel tasks and tool ecosystems, then rewrites them into intentionally underspecified instructions. This compels agents to actively query users for missing details. When handling synthetic tasks, an LLM-based user simulator provides user-private information, while a mock tool system delivers stable tool responses. For rewards, task-level rubrics are constructed based on required subgoals, user-agent interactions, and forbidden behaviors. Across 14 challenging datasets in math, search, and tool use, models trained on our synthetic data achieve substantial gains, with small models outperforming larger baselines.