sulabhkatiyar/en-indic-translate-26b
The sulabhkatiyar/en-indic-translate-26b is a 26 billion parameter Mixture-of-Experts (MoE) model, fine-tuned from Google's Gemma-4-26B-A4B-it, designed for English to 11 Indic language translation. It excels at translating complex scientific documents by preserving LaTeX formulas, code blocks, and document structure, maintaining coherence over long texts. This model specifically targets high-quality translation of structured content, outperforming baselines in math preservation and structural integrity.
Loading preview...
En-Indic Translate 26B: Specialized Document Translation
This model, sulabhkatiyar/en-indic-translate-26b, is a 26 billion parameter Mixture-of-Experts (MoE) model, fine-tuned from google/gemma-4-26B-A4B-it. It is specifically engineered for high-quality English to 11 Indic language translation, with a particular focus on complex, structured documents.
Key Capabilities
- Multilingual Translation: Translates English into Assamese, Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu.
- Structure Preservation: Uniquely designed to preserve LaTeX formulas, code blocks, and overall document structure during translation.
- Coherence over Long Texts: Maintains contextual coherence across thousands of tokens, crucial for scientific and technical documents.
- Robust Performance: Achieves significantly higher scores than the
sarvamai/sarvam-translatebaseline in metrics like math preservation (0.980 vs 0.364), structure integrity (0.991 vs 0.607), and renderability (95.8% vs 84.6%). - Low Degeneracy: Exhibits a low degeneracy (loop) rate of 2.2%, indicating stable and reliable output generation.
Ideal Use Cases
- Scientific Document Translation: Perfect for translating research papers, technical manuals, and academic texts where mathematical equations and code snippets must remain intact.
- Technical Content Localization: Suitable for localizing software documentation, engineering specifications, and other structured content into Indic languages.
- High-Fidelity Translation: When preserving the exact layout and non-textual elements of a document is critical, this model offers superior performance compared to general-purpose translation models.