linglingdan/DRIFT-8B-Physics
DRIFT-8B-Physics is an 8 billion parameter causal language model fine-tuned from Qwen/Qwen3-8B by linglingdan, specifically optimized for physics-related tasks. It features a context length of 32768 tokens, leveraging the Qwen3ForCausalLM architecture. This model is designed to provide specialized assistance in scientific domains, particularly physics, building upon its Qwen3 base.
Loading preview...
Model Overview
linglingdan/DRIFT-8B-Physics is an 8 billion parameter language model, fine-tuned from the Qwen/Qwen3-8B base model. It utilizes the Qwen3ForCausalLM architecture and supports a substantial context length of 32768 tokens (with a configured max_position_embeddings of 40960).
Key Characteristics
- Base Model: Fine-tuned from
Qwen/Qwen3-8B. - Architecture: Employs
Qwen3ForCausalLM. - Precision: Operates in
bfloat16for efficient computation. - Context Window: Features a large context window, making it suitable for processing extensive physics-related texts and problems.
- Research Origin: Developed as part of the DRIFT research project, detailed in the academic paper arXiv:2606.30345.
Usage and Limitations
This model is intended for applications requiring specialized knowledge in physics. It comes with a default generation_config.json (temperature=0.6, top_k=20, top_p=0.95) which can be adjusted for specific deployment needs. Users should be aware of potential limitations, including possible hallucination of tool calls or invalid arguments, and that output quality can depend on the serving template and tool schema formatting. Safety, bias, and domain-specific failure modes are not fully documented.