---
library_name: transformers
pipeline_tag: text-generation
tags:
- qwen3
- tool-calling
- bfcl
- agentic-rl
- progress-reward
- online-synthesis
---

<div align="center">

<img src="assets/toolweave-mark.svg" alt="ToolWeave mark" width="128">

<h1>ToolWeave · Stage 3</h1>

<p><strong>🧭 Boundary-Guided Online Reinforcement Learning</strong></p>
<p>Verified online data synthesis for multi-turn tool-calling agents.</p>

<p>
  <a href="https://github.com/Muradil-mamat-211/ToolWeave">🧵 Project</a>
</p>

</div>

## 🧭 At a glance

| Field | Details |
|---|---|
| 🧠 Base family | Qwen3-4B-Instruct |
| 🪜 Curriculum stage | Stage 3 — Boundary-Guided Online Reinforcement Learning |
| 🧱 Starting point | ToolWeave Stage 2 update 25 |
| 🎛️ Training signal | Verified online data synthesis + multi-turn Progress Reward |
| ✅ Release status | Final ToolWeave Stage 3 model |

ToolWeave Stage 3 expands multi-turn tool-use learning through capability-boundary detection, verified online data synthesis, strict execution and semantic validation, dynamic replay, and combined global/local tool-call credit.

## 📊 Stage 3 evaluation

The final ToolWeave Stage 3 checkpoint was evaluated on the canonical balanced 400-row held-in set: 100 entries each from Base, Missing Function, Missing Parameter, and Long Context. These values are complete-entry BFCL Multi-Turn accuracies, not the training-time Progress Reward (`R_P`).

| Model | Overall | Base | Missing Function | Missing Parameter | Long Context | Correct entries |
|---|---:|---:|---:|---:|---:|---:|
| **ToolWeave Stage 3** | **48.50** | **56.00** | **50.00** | **42.00** | **46.00** | **194 / 400** |

Because the four categories are balanced, the overall score is their unweighted mean and the complete-entry accuracy over all 400 entries:

`(56.00 + 50.00 + 42.00 + 46.00) / 4 = 48.50`

## 🚀 Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "muradil211/ToolWeave_stage3"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
```

Tool-use inference requires the model's function schemas and the Qwen3-compatible tool-call format.

## 🔗 Links

- [🧵 ToolWeave project](https://github.com/Muradil-mamat-211/ToolWeave)