LiquidAI/LFM2-1.2B-Longevity
LFM2-1.2B-Longevity is a 1.2 billion parameter language model developed jointly by Insilico Medicine and Liquid AI, featuring a hybrid Liquid architecture with multiplicative gates and short convolutions. This model is specifically fine-tuned for interpreting heterogeneous aging biology data, including multi-omics and clinical information, and operates with a 32,768-token context length. It is designed for research in aging biology and omics interpretation, providing domain-adapted insights into complex biological datasets.
Loading preview...
Overview
LFM2-1.2B-Longevity is a 1.2 billion parameter model from the Longevity-LLM family, developed by Insilico Medicine and Liquid AI. It is a compact, domain-adapted language model specifically designed for interpreting heterogeneous aging biology data. The model is built on a hybrid Liquid architecture, incorporating multiplicative gates and short convolutions, and supports a substantial context length of 32,768 tokens.
Key Capabilities
- Domain Adaptation: Fine-tuned on aging-related multi-omics and clinical data, making it highly specialized for this field.
- Architecture: Utilizes a unique Hybrid Liquid model architecture for efficient processing.
- Context Length: Features a long context window of 32,768 tokens, enabling the processing of extensive biological data.
- Training Data: Trained on the L-LLM corpus, which spans aging biology, detailed in the LongevityBench dataset.
- Chat Template: Employs a ChatML-like format with a dynamic-thinking template for flexible interaction.
Intended Use and Limitations
This model is intended for research in aging biology and omics interpretation. It provides model predictions that should be experimentally validated and not used as clinical advice. Its performance is optimized for the data modalities present in its training corpus.