Results for "task-specific"

Reusing knowledge from a source task/domain to improve learning on a target task/domain, typically via pretrained models.

Updating a pretrained model’s weights on task-specific data to improve performance or adapt style/behavior.

Assigning category labels to images.

Reward only given upon task completion.

Training one model on multiple tasks simultaneously to improve generalization through shared structure.

Achieving task performance by providing a small number of examples inside the prompt without weight updates.

Measure of consistency across labelers; low agreement indicates ambiguous tasks or poor guidelines.

Pixel-level separation of individual object instances.

Pixel-wise classification of image regions.

Decomposing goals into sub-tasks.

Maximizing reward without fulfilling real goal.

Task instruction without examples.

Breaking tasks into sub-steps.

Fine-tuning on (prompt, response) pairs to align a model with instruction-following behaviors.

PEFT method injecting trainable low-rank matrices into layers, enabling efficient fine-tuning.

Tendency for agents to pursue resources regardless of final goal.

The field of building systems that perform tasks associated with human intelligence—perception, reasoning, language, planning, and decision-making—via algori...

Tradeoffs between many layers vs many neurons per layer.

A high-capacity language model trained on massive corpora, exhibiting broad generalization and emergent behaviors.

Combining signals from multiple modalities.

Detects trigger phrases in audio streams.

A scalar measure optimized during training, typically expected loss over data, sometimes with regularization terms.

Automatically learning useful internal features (latent variables) that capture salient structure for downstream tasks.

A dataset + metric suite for comparing models; can be gamed or misaligned with real-world goals.

Attacks that manipulate model instructions (especially via retrieved content) to override system goals or exfiltrate data.

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

Assigning labels per pixel (semantic) or per instance (instance segmentation) to map object boundaries.

Identifying speakers in audio.

Models trained to decide when to call tools.

Model exploits poorly specified objectives.