metadata
language: en
tags:
- code
- algorithms
- competitive-programming
- multi-label-classification
- codebert
datasets:
- xCodeEval
metrics:
- f1
- precision
- recall
library_name: transformers
pipeline_tag: text-classification
CodeBERT Algorithm Tagger
A fine-tuned CodeBERT model for multi-label classification of algorithmic problems from competitive programming platforms like Codeforces.
Model Description
This model predicts algorithmic tags/categories for competitive programming problems based on their problem descriptions and solution code.
Supported Tags:
- math
- graphs
- strings
- number theory
- trees
- geometry
- games
- probabilities
Training Data
- Dataset: xCodeEval (Codeforces problems)
- Training examples: 2,147 problems (filtered for focus tags)
- Test examples: 531 problems
- Source: Problems and solutions from Codeforces platform
Model Architecture
- Input: Concatenated problem description and solution code
- Encoder: CodeBERT (RoBERTa-based architecture)
- Output: 8-dimensional binary classification (one per tag)
Usage
Installation
pip install transformers torch