LOGOS-Hub/LOGOS-pretrain-3B
LOGOS-Hub/LOGOS-pretrain-3B is a 3.2 billion parameter autoregressive Transformer model developed by LOGOS-Hub, designed as the first multi-domain generative framework for natural sciences. It utilizes a unified scientific grammar to encode diverse scientific objects like proteins, molecules, and materials into token sequences. This model excels at generation, prediction, and design tasks across various scientific domains, operating directly on domain-native representations without explicit 3D geometry.
Loading preview...
LOGOS-pretrain-3B: A Unified Generative Model for Natural Sciences
LOGOS (Language Of Generative Objects in Science) is a 3.2 billion parameter autoregressive Transformer model, part of a family ranging from 1B to 8B parameters. It introduces the first multi-domain generative framework built upon a unified scientific grammar. This grammar encodes heterogeneous scientific objects, such as proteins, antibodies, small molecules, chemical reactions, and materials, along with their spatial interactions, into a common discrete token space.
Key Capabilities
- Unified Scientific Grammar: Employs a shared representational interface for diverse scientific objects and their relationships.
- Multi-Domain Functionality: A single model handles generation, prediction, and design across various natural science domains.
- No Explicit 3D Geometry: Captures complex spatial contact and constraint patterns through tokenized representations, eliminating the need for geometric neural networks.
- Consistent Pre-training & Downstream Alignment: Ensures formal consistency between pre-training objectives and downstream task goals within the grammar space.
Good For
LOGOS-pretrain-3B is suitable for a wide array of scientific tasks, demonstrating competitive performance in areas such as:
- Drug Discovery: Interaction-aware ligand design for protein binding pockets.
- Structural Biology: Identifying ligand-binding sites from protein sequences.
- Chemistry: Retrosynthesis prediction.
- Materials Science: Unconditional generation of novel materials.
- Protein Engineering: Editing protein sequences for functional improvements.
- Immunology: Designing complementarity-determining regions (CDRs) for antibody engineering.