bbhqwen3 / README.md
sumink's picture
Create README.md
d807519 verified
๐Ÿš€ Qwen2.5-3B Fine-Tuned on BBH (Dyck Languages) - Model Card
๐Ÿ“Œ Model Overview
Model Name: Qwen2.5-3B Fine-Tuned on BBH (Dyck Languages)
Base Model: Qwen2.5-3B-Instruct
Fine-Tuned Dataset: BBH (BigBench Hard) - Dyck Languages
Task: Causal Language Modeling (CLM)
Fine-Tuning Objective: Improve performance on Dyck language sequence completion (correctly closing nested parentheses and brackets)
๐Ÿ“Œ Dataset Information
This model was fine-tuned on the Dyck Languages subset of the BigBench Hard (BBH) dataset.
Dataset characteristics:
Task Type: Sequence completion of balanced parentheses
Input Format: A sequence of open parentheses, brackets, or braces with missing closing elements
Target Labels: The correct sequence of closing parentheses, brackets, or braces
Example:
Input: Complete the rest of the sequence, making sure that the parentheses are closed properly. Input: [ [
Target: ] ]
This dataset evaluates a modelโ€™s ability to correctly complete structured sequences, which is crucial for programming language syntax, formal language understanding, and symbolic reasoning.