GenesisFunc: Multi-Agent Data Generation for Accurate and Generalizable Function-Calling
Quick Answer
GenesisFunc introduces an automated multi-agent pipeline for generating high-quality function-calling (FC) training data, outperforming existing models in both in-domain performance and out-of-domain generalization.
Quick Take
GenesisFunc introduces an automated pipeline for generating high-quality function-calling (FC) training data, outperforming existing models in both in-domain performance and out-of-domain generalization. Fine-tuned on an 8B LLM, it demonstrates comparable FC capabilities to leading API-based models, addressing challenges of diversity and quality in synthetic data generation.
Key Points
- GenesisFunc employs a multi-agent framework for diverse dialogue generation.
- Fine-tuned 8B LLM outperforms similar-sized open-source models in FC tasks.
- Demonstrates strong scalability potential across various downstream tools.
- Multi-stage evaluation system enhances data accuracy and quality.
- Addresses challenges of limited diversity and weak quality control in synthetic data.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Large Language Models (LLMs) extend their capabilities through function-calling (FC), which relies on training data with high quality, diversity, and broad coverage of scenario. However, obtaining and annotating real function-calling data is challenging, while synthetic data from existing pipelines often suffers from unreliable APIs, limited tool scalability, insufficient diversity, and weak quality control. To address these, we present GenesisFunc, an automated pipeline for generating FC training data. Starting from reliable tools in widely used public benchmarks, our GenesisFunc employs a multi-agent framework to support a dialogue generation system that produces conversations spanning diverse scenarios, while maintaining both diversity and quality throughout the process. The accuracy of the data is further reinforced through a multi-stage evaluation system. We fine-tune an 8B LLM on the synthetic dataset and show through extensive experiments that it outperforms similarly sized open-source models in in-domain FC performance and out-of-domain generalization, while reaching FC capabilities comparable to some of the latest API-based models. In addition, our method demonstrates strong potential to scale effectively across downstream tools, underscoring its real-world applicability.
| Comments: | Accepted by ACL 2026 Main |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2605.28835 [cs.CL] |
| (or arXiv:2605.28835v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.28835 arXiv-issued DOI via DataCite |
Submission history
From: Haoxiang Xu [view email]
[v1]
Fri, 10 Apr 2026 14:02:03 UTC (1,707 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →Letting the Data Speak: Extracting Keywords from Crowdsourced Collections with AI
The study evaluates three NLP approaches—Named Entity Recognition, Keyword Extraction, and Topic Modelling—using the Their Finest Hour Online Archive to automate keyword extraction from crowdsourced WWII collections. Findings suggest that while NLP methods show promise, no single approach is sufficient, and ethical considerations in automated keyword extraction are crucial for responsible stewardship.