Explore ML tasks

Generate text given a prompt, including open-ended generation and conditional generation.

Classify text into predefined categories such as sentiment, topic, or intent.

Label individual tokens in a sequence — used for NER, POS tagging, and chunking.

Extract or generate answers to questions based on a given context passage.

Produce a shorter version of a document while preserving key information.

Convert text from one natural language to another.

Predict masked tokens in a sequence — the core pre-training objective for BERT-style models.

Compute semantic similarity scores between pairs of sentences.

Extract dense vector embeddings from text for downstream tasks like search and clustering.

Classify text into categories never seen during training using natural language labels.

Assign one or more labels to an input image from a fixed set of categories.

Locate and classify multiple objects within an image with bounding boxes.

Assign a class label to every pixel in an image (semantic or instance segmentation).

Predict per-pixel depth maps from a single RGB image.

Transform or enhance an input image — style transfer, super-resolution, inpainting.

Generate photorealistic or artistic images from natural language prompts.

Generate descriptive text captions or OCR output from images.

Label video clips with action or event categories.

Transcribe spoken audio to text — the core task for voice assistants and captioning.

Convert written text to natural-sounding speech audio.

Classify audio clips into categories such as music genre, speaker, or sound event.

Transform audio signals — noise reduction, speech enhancement, source separation.

Answer natural language questions about the content of an image.

Extract answers from structured documents such as PDFs, forms, and charts.

Animate a still image into a short video clip.

Generate video from natural language descriptions.

Predict categorical labels from structured tabular features.

Predict continuous numerical targets from structured tabular features.

Train agents to maximize rewards through interaction with an environment.

Learn from graph-structured data — node classification, link prediction, graph classification.