lhpku20010120/qwen3-4b-k12kgraph

TEXT GENERATIONConcurrent Unit Cost:1Model Size:4BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Jul 28, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

lhpku20010120/qwen3-4b-k12kgraph is a 4 billion parameter Qwen3ForCausalLM model, fine-tuned by lhpku20010120 using full-parameter supervised fine-tuning on the K12-Train dataset. This model is specifically designed for K-12 educational question answering and knowledge-intensive reasoning, supporting both Chinese and English. It excels in educational contexts derived from knowledge graphs, providing specialized responses for academic queries.

Loading preview...

Model Overview

lhpku20010120/qwen3-4b-k12kgraph is a 4 billion parameter language model built upon the Qwen3-4B-Base architecture. It has undergone full-parameter supervised fine-tuning (SFT) using the K12-Train dataset, which is associated with the K12-KGraph project. This specialization makes it particularly adept at handling educational content.

Key Capabilities

  • Specialized Educational QA: Fine-tuned for K-12 educational question answering, making it suitable for academic support systems.
  • Knowledge-Intensive Reasoning: Designed to perform knowledge-intensive reasoning, especially within domains informed by knowledge graphs.
  • Multilingual Support: Supports both Chinese and English languages.
  • Standalone Checkpoint: Provided as a complete Transformers checkpoint, eliminating the need to download the base model separately.

Intended Use Cases

  • K-12 Educational Research: Ideal for research into educational question answering and the evaluation of models trained with knowledge-graph-derived data.
  • Academic Support Systems: Can be integrated into applications requiring specialized responses to K-12 academic queries.

Limitations

  • The model may exhibit biases or errors present in its base model or fine-tuning data.
  • Generated answers might be incorrect, incomplete, or subject to hallucination; it should not be considered an authoritative source for critical educational decisions.
  • Performance is optimized for K-12 educational content and may degrade significantly outside this specific domain.