Results for "API-only"

Models accessible only via service APIs.

Reconstructing a model or its capabilities via API queries or leaked artifacts.

Reward only given upon task completion.

Forcing predictable formats for downstream systems; reduces parsing errors and supports validation/guardrails.

Routes inputs to subsets of parameters for scalable capacity.

Capabilities that appear only beyond certain model sizes.

When a model fits noise/idiosyncrasies of training data and performs poorly on unseen data.

Activation max(0, x); improves gradient flow and training speed in deep nets.

Rules and controls around generation (filters, validators, structured outputs) to reduce unsafe or invalid behavior.

Training across many devices/silos without centralizing raw data; aggregates updates, not data.

Protecting data during network transfer and while stored; essential for ML pipelines handling sensitive data.

Automated testing and deployment processes for models and data workflows, extending DevOps to ML artifacts.

Techniques that fine-tune small additional components rather than all weights to reduce compute and storage.

Identifying and localizing objects in images, often with confidence scores and bounding rectangles.

Attention mechanisms that reduce quadratic complexity.

Models evaluating and improving their own outputs.

Formal framework for sequential decision-making under uncertainty.

Pixel-level separation of individual object instances.

Models whose weights are publicly available.

Prompt augmented with retrieved documents.

Enables external computation or lookup.

Designing AI to cooperate with humans and each other.

Inferring and aligning with human preferences.

Learning only from current policy’s data.

Decisions dependent on others’ actions.