ahammad115566/qwen-smeft

TEXT GENERATIONConcurrent Unit Cost:1Model Size:8BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 31, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Featherless Exclusive Cold

The ahammad115566/qwen-smeft model is a domain-adapted large language model fine-tuned from Qwen3-8B by Ahmed Hammad and Veronica Sanz. It is specifically optimized for tasks related to Standard Model Effective Field Theory (SMEFT), including operator reasoning, EFT basis translation, and physics-aware scientific dialogue. This model excels at structured theoretical question answering within the particle physics domain, leveraging a curated corpus of SMEFT and high-energy physics literature.

Loading preview...

Overview

This model, developed by Ahmed Hammad and Veronica Sanz, is a domain-adapted large language model specifically fine-tuned for Standard Model Effective Field Theory (SMEFT). It is built upon the Qwen3-8B base model and was fine-tuned using LoRA with a curated corpus of SMEFT and high-energy physics preprints. The training focused on instruction-following scientific question answering within this specialized domain.

Key Capabilities

  • SMEFT operator reasoning: Understands and processes complex SMEFT operators.
  • EFT basis translation: Capable of translating between different Effective Field Theory bases.
  • Physics-aware scientific dialogue: Engages in technical discussions grounded in particle physics principles.
  • Literature-style technical explanation: Generates explanations in a style consistent with scientific literature.
  • Structured theoretical question answering: Provides structured answers to theoretical physics questions.

Limitations and Considerations

  • Domain-locked: This model is intentionally designed for SMEFT and is not suitable for general-purpose tasks.
  • Potential for hallucination: May occasionally hallucinate operator identities.
  • Uneven coverage: Due to 3620 training examples, coverage of the SMEFT operator space might be uneven, potentially leading to less reliable answers for rare operators or non-Warsaw bases.

Usage Note

When using Qwen3's native enable_thinking toggle, set it to False to avoid conflicts with this model's custom structured prompt format for reasoning.