OrdenWills/LFM2.5-350M-home-assistant-sft
The OrdenWills/LFM2.5-350M-home-assistant-sft is a 350 million parameter smart home automation model, fine-tuned from LiquidAI's LFM2.5-350M. It is specifically designed to interpret natural language commands and generate structured tool calls for controlling smart home devices like lights, doors, thermostats, and media players, with full awareness of current device states. This model excels at complex disambiguation, pronoun resolution, and handling multi-device compound commands, achieving 98.5% accuracy across 59 specific categories.
Loading preview...
Model Overview
The OrdenWills/LFM2.5-350M-home-assistant-sft is a specialized 350 million parameter language model, fine-tuned from LiquidAI/LFM2.5-350M, designed for smart home automation. Its core function is to translate natural language commands into precise, structured tool calls for controlling various home devices, while maintaining full awareness of their current states.
Key Capabilities
- Advanced Device Disambiguation: Intelligently resolves ambiguous commands (e.g., "Turn off the TV") by considering device states, room context, and connected devices.
- State Awareness: Handles commands based on current device states, such as gracefully responding when lights are already off or dynamically acting on devices that need action (e.g., "Off everything").
- Complex Command Handling: Supports media playback, pronoun resolution, relative state clauses ("Close the door that is open"), and multi-device compound commands, emitting parallel tool calls without truncation.
- Robust Tool Calling: Utilizes a 10-tool schema for controlling lights, doors, thermostats, scenes, TVs, fans, and speakers, with strict syntactic triggers to prevent JSON bleed.
- High Accuracy: Achieves an overall accuracy of 98.5% across 59 highly specific edge-case and logic-stress categories in evaluation.
Training and Features
This model was fine-tuned on a 160,000-example state-aware synthetic dataset ("V14"). Notable training features include positive-only "Think Traces" to reduce hallucination, rigorous syntactic triggers for tool call boundaries, and compound count enforcement to prevent truncation of multi-action requests.
Use Cases
This model is ideal for developers building intelligent smart home assistants that require precise, context-aware control over connected devices. It's particularly suited for applications needing robust natural language understanding for home automation tasks, offering a highly reliable and efficient solution for converting voice or text commands into actionable device controls.