# Synthesize and Reward -- Reinforcement Learning for Multi-Step Tool Use in Live Environments

arXiv: https://arxiv.org/abs/2606.03892 · HTML: https://mlsanity.com/paper/2606.03892/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use.
- Builds on 14 of its 26 references that are in the map.
- First posted on arXiv 2 Jun 2026.

## Placed in

- LLM agents and tool use / [Reinforcement learning for LLM agents](https://mlsanity.com/agents/reinforcement-learning-for-llm-agents/)

## Builds on (most cited first, 14 of 14)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [τ2-Bench: Evaluating Conversational Agents in a Dual-Control Environment](https://mlsanity.com/paper/2506.07982/) (2025)
- [ReTool: Reinforcement Learning for Strategic Tool Use in LLMs](https://mlsanity.com/paper/2504.11536/) (2025)
- [xLAM: A Family of Large Action Models to Empower AI Agent Systems](https://mlsanity.com/paper/2409.03215/) (2024)
- [T-Eval: Evaluating the Tool Utilization Capability of Large Language Models Step by Step](https://mlsanity.com/paper/2312.14033/) (2023)
- [ComplexFuncBench: Exploring Multi-Step and Constrained Function Calling under Long-Context Scenario](https://mlsanity.com/paper/2501.10132/) (2025)
- [When2Call: When (not) to Call Tools](https://mlsanity.com/paper/2504.18851/) (2025)
- [Magnet: Multi-turn Tool-use Data Synthesis and Distillation via Graph Translation](https://mlsanity.com/paper/2503.07826/) (2025)
- [StepTool: Enhancing Multi-Step Tool Usage in LLMs via Step-Grained Reinforcement Learning](https://mlsanity.com/paper/2410.07745/) (2024)
- [Tool Zero: Training Tool-Augmented LLMs via Pure RL from Scratch](https://mlsanity.com/paper/2511.01934/) (2025)
- [Tool-R1: Sample-Efficient Reinforcement Learning for Agentic Tool Use](https://mlsanity.com/paper/2509.12867/) (2025)
- [ToolACE-MT: Non-Autoregressive Generation for Agentic Multi-Turn Interaction](https://mlsanity.com/paper/2508.12685/) (2025)
- [Simulating Complex Multi-Turn Tool Calling Interactions in Stateless Execution Environments](https://mlsanity.com/paper/2601.19914/) (2026)
- [Training LLMs for Multi-Step Tool Orchestration with Constrained Data Synthesis and Graduated Rewards](https://mlsanity.com/paper/2603.24709/) (2026)

## Built on by (most cited first, 1 of 1)

- [Davinci-Tool: Scaling Agentic Tool-Use Supervision with Verifiable Executable Environments](https://mlsanity.com/paper/2610.16867/) (2026)

## Abstract (arXiv)

Training LLMs to orchestrate multi-step tool calls is held back by three coupled obstacles: realistic stateful execution environments are costly to build, synthetic training queries are often detached from the server's actual state (so the generated tool calls fail to execute), and recall-based RL rewards incentivize verbose tool-calling patterns. We present PROVE (Programmatic Rewards On Verified Environments), a framework with three contributions: (1) a library of 20 stateful MCP (Model Context Protocol) servers exposing 343 tools, enabling live-execution RL training with session-scoped state isolation; (2) a state-machine data synthesis pipeline that generates multi-turn tool-call trajectories grounded in live-sampled server state, so generated queries reference entities that actually exist; and (3) a multi-component programmatic reward with an adaptive efficiency penalty that counters the verbosity incentive of recall-based rewards. We train four models (Qwen3-4B, Qwen3-8B, Qwen2.5-7B, Granite-4.1-8B) with GRPO on the resulting ~13K training examples. On BFCL Multi-Turn, tau2-bench, and T-Eval, PROVE yields improvements of up to +10.2, +6.8, and +6.5 points respectively, demonstrating that this framework yields consistent gains on multi-step tool orchestration across two model families.