Prompt-Driven Tool-Calling for Lightweight Open Source LLMs (AIS2C2 2025 Paper)

Researchers at AIS2C2 2025 presented a prompt-driven tool-calling framework enabling lightweight open-source LLMs to perform complex multi-step tasks, reducing reliance on large proprietary models and lowering compute requirements.

Hi Hugging Face community I’m excited to share my recent research work published at AIS2C2 2025. Paper: Prompt-Driven Tool-Calling for Lightweight Open Source LLMs This research explores how lightweight open-source LLMs can perform complex tasks using a prompt-driven tool-calling framework , without relying on large proprietary models. Problem Statement Most small LLMs struggle with: - Multi-step reasoning - Tool usage APIs, calculators, search tools - Complex task decomposition Large models solve this, but they are expensive and not always deployable. Proposed Idea We introduce a prompt-driven approach that enables LLMs to: - Decide when to use tools - Select appropriate tools dynamically - Chain multiple reasoning steps - Reduce dependency on large models Key Benefits - Works with lightweight open-source LLMs - Reduces compute requirements - Enables AI agent-like behavior - Practical for real-world deployment Paper Link https://www.aiscindia.co.in/wp-content/uploads/2026/06/ilovepdf merged-4.pdf https://www.aiscindia.co.in/wp-content/uploads/2026/06/ilovepdf merged-4.pdf Looking for Feedback I would love feedback from the community on: - Better tool-calling strategies - Improvements for lightweight agents - Open-source implementations Thank you