Classification

Learn a classifier over sentence embeddings to assign labels to individual sentences.

Metric: F1

Multi-label Classification

Predict multiple labels for each input text using a classifier trained on embeddings.

Metric: F1

Pair Classification

Predict a binary relationship between two sentences based on their embedding similarity.

Metric: Average Precision (AP)

Semantic Textual Similarity (STS)

Score sentence pair similarity using embedding distance metrics.

Metric: Cosine Similarity

Clustering

Group semantically similar texts into clusters using k-means over embeddings.

Metric: V-measure

Bitext Mining

Identify cross-lingual translation pairs via nearest-neighbour embedding retrieval.

Metric: F1

Retrieval

Retrieve relevant documents for a query using embedding similarity search.

Metric: NDCG@10

Instruction Retrieval

Retrieve documents using queries enriched with natural language relevance instructions.

Metric: NDCG@5

Reranking

Improve relevance ranking by reordering candidates using embedding similarity.

Metric: Mean Average Precision (MAP)