kiki-ailab/Llama3.2-1B-Instruct-KAI

Hugging Face
TEXT GENERATIONPricing:Input $0.108 / Output $0.804Concurrent Unit Cost:1Model Size:1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Feb 19, 2025License:otherArchitecture:Transformer Featherless Exclusive Warm

Llama3.2-1B-Instruct-KAI by kiki-ailab is a 1 billion parameter instruction-tuned causal language model, fine-tuned from the Llama3.2 base model. It is specifically optimized for Vietnamese language understanding and generation tasks, including reading comprehension, information extraction, question answering, and summarization. This model demonstrates significant performance improvements on Vietnamese benchmarks like VMLU, making it suitable for applications requiring strong Vietnamese NLP capabilities.

Loading preview...

Llama3.2-1B-Instruct-KAI: Optimized for Vietnamese NLP

Llama3.2-1B-Instruct-KAI is a 1 billion parameter instruction-tuned model developed by kiki-ailab, built upon the Llama3.2 base architecture. This model is part of a collection specifically fine-tuned to excel in Vietnamese language understanding and generation tasks.

Key Capabilities

  • Vietnamese Language Proficiency: Optimized for tasks such as reading comprehension, information extraction, question answering, and summarization in Vietnamese.
  • Enhanced Performance: Demonstrates substantial improvements over its base model on various Vietnamese benchmarks, including VMLU, ViSquad, ViDrop, and ViDialog. For instance, on VMLU, it achieves 50.5, an increase of 12.9 points over the base Llama3.2-1B-Instruct.
  • Instruction Following: Designed to follow instructions effectively for diverse NLP applications.

Good For

  • Vietnamese-centric Applications: Ideal for developers building applications that require robust Vietnamese natural language processing.
  • Question Answering Systems: Excels in extracting answers from provided text in Vietnamese.
  • Summarization Tasks: Capable of generating concise summaries of Vietnamese documents.
  • Information Extraction: Useful for identifying and extracting key information from Vietnamese content.

Limitations

  • May exhibit hallucinations on culturally specific content.
  • Primarily focused on Vietnamese language, and may not perform optimally for specialized technical domains or other languages.

Finetunes
2 models