Models Qwen 2.51b5

Salesforce/xLAM-2-1b-fc-r

Warm
Public
1.5B
BF16
131072
Mar 27, 2025
License: cc-by-nc-4.0
Hugging Face
Overview

xLAM-2-1b-fc-r: A Large Action Model for AI Agents

The Salesforce/xLAM-2-1b-fc-r is a 1.5 billion parameter model from the xLAM-2 series, developed by Salesforce AI Research. This model is specifically designed as a Large Action Model (LAM), acting as the "brain" for AI agents by translating user intentions into executable actions. It is built upon advanced data synthesis, processing, and training pipelines, notably utilizing the APIGen-MT framework which generates high-quality training data through simulated agent-human interactions.

Key Capabilities & Features

  • Multi-turn Conversation: Optimized for complex, multi-turn dialogues, enabling more natural and effective agent interactions.
  • Function-Calling (fc-r suffix): Fine-tuned for robust function-calling tasks, allowing AI agents to autonomously plan and execute tasks by calling external tools.
  • State-of-the-Art Performance: Achieves leading results on the BFCL leaderboard and Ļ„-bench benchmarks, outperforming larger frontier models in multi-turn scenarios and consistency.
  • Research Release: Designated as a research release, focusing on advancing AI agent capabilities.
  • vLLM Integration: Features refined chat templates and seamless vLLM integration for efficient and high-throughput inference.
  • Context Length: Supports a context length of 131,072 tokens, with a default of 32k for Qwen-2.5-based models, extendable via techniques like YaRN.

Ideal Use Cases

This model is particularly well-suited for research and development in:

  • AI Agent Systems: Building and enhancing AI agents that require advanced decision-making and autonomous task execution.
  • Automated Workflows: Developing systems that automate complex workflows across diverse domains through intelligent action planning.
  • Function-Calling Applications: Implementing applications that leverage precise and consistent function-calling capabilities in conversational AI.

This model is released for research purposes only, and users are encouraged to consider ethical implications and conduct thorough evaluations for specific downstream applications.