The Salesforce xLAM-2-3b-fc-r is a 3.1 billion parameter Large Action Model (LAM) developed by Salesforce, designed to translate user intentions into executable actions for AI agents. This model excels in multi-turn conversation and tool usage, trained using the novel APIGen-MT framework for high-quality data generation. It achieves state-of-the-art performance on BFCL and τ-bench benchmarks, outperforming larger models in function-calling and agentic capabilities, making it ideal for automating complex workflows.
Loading preview...
Salesforce xLAM-2-3b-fc-r: Advanced Large Action Model
The Salesforce xLAM-2-3b-fc-r is part of the new xLAM-2 series, a family of Large Action Models (LAMs) developed by Salesforce AI Research. These models are engineered to act as the "brains of AI agents," autonomously planning and executing tasks by translating user intentions into actionable steps. This specific model, with 3.1 billion parameters and a 32k context length (extendable to 128k), is fine-tuned for function calling (-fc) and is released for research purposes (-r).
Key Capabilities & Differentiators
- Multi-turn Conversation & Tool Usage: The xLAM-2 series marks a significant advancement in handling complex multi-turn dialogues and effectively utilizing external tools.
- APIGen-MT Training: Trained using the novel APIGen-MT framework, which generates high-quality data through simulated agent-human interactions, ensuring robust performance.
- State-of-the-Art Benchmarks: Achieves superior performance on the BFCL (Berkeley Function-Calling Leaderboard) and τ-bench benchmarks, outperforming frontier models like GPT-4o and Claude 3.5 in agentic capabilities, even with its smaller size.
- Consistency: Demonstrates exceptional consistency across trials in multi-turn scenarios.
- Seamless Integration: Features a refined chat template and vLLM integration for easier deployment and building advanced AI agents.
Ideal Use Cases
- AI Agent Development: Building sophisticated AI agents that require advanced decision-making and autonomous task execution.
- Function Calling Applications: Scenarios demanding precise and reliable function calling, where the model needs to interact with external APIs or tools.
- Research & Development: Exploring advanced multi-turn conversational AI and agentic systems, particularly for evaluating novel training methodologies like APIGen-MT.