AnirbanSaha/qwen2.5-14b-tlink-full-autoregressive
AnirbanSaha/qwen2.5-14b-tlink-full-autoregressive is a 14.8 billion parameter language model, fine-tuned from Qwen/Qwen2.5-14B-Instruct, specifically for temporal relation classification. It utilizes full-parameter fine-tuning on the fahmidiqbal/tlink-classification dataset to identify relationships like BEFORE, AFTER, OTHER, or NONE between temporal expressions. This model is optimized for tasks requiring precise temporal understanding and achieves an accuracy of 0.8281 on its test set.
Loading preview...
Overview
AnirbanSaha/qwen2.5-14b-tlink-full-autoregressive is a 14.8 billion parameter language model derived from the Qwen/Qwen2.5-14B-Instruct base model. It has undergone full-parameter fine-tuning to specialize in temporal relation classification, a task crucial for understanding event sequences and timelines in text. The model was trained using the fahmidiqbal/tlink-classification dataset.
Key Capabilities
- Temporal Relation Classification: Accurately identifies temporal relationships between entities, categorizing them as BEFORE, AFTER, OTHER, or NONE.
- Full-Parameter Fine-Tuning: Leverages comprehensive fine-tuning to achieve specialized performance in its target domain.
- High Accuracy: Achieves a test accuracy of 0.8281, with a macro-F1 of 0.8272 and weighted-F1 of 0.8285, demonstrating robust performance in temporal linking.
Good for
- Event Ordering and Sequencing: Ideal for applications requiring the precise ordering of events or actions within documents.
- Information Extraction: Useful for extracting structured temporal information from unstructured text.
- Natural Language Understanding (NLU): Enhances NLU systems by providing a strong component for temporal reasoning.