Results for "perception input"

Networks using convolution operations with weight sharing and locality, effective for images and signals.

Attention where queries/keys/values come from the same sequence, enabling token-to-token interactions.

Architecture based on self-attention and feedforward layers; foundation of modern LLMs and many multimodal models.

Studying internal mechanisms or input influence on outputs (e.g., saliency maps, SHAP, attention analysis).

Local surrogate explanation method approximating model behavior near a specific input.

Models that process or generate multiple modalities, enabling vision-language tasks, speech, video understanding, etc.

A narrow hidden layer forcing compact representations.

Allows model to attend to information from different subspaces simultaneously.

Routes inputs to subsets of parameters for scalable capacity.

Probabilistic graphical model for structured prediction.

Detects trigger phrases in audio streams.

Control without feedback after execution begins.

Running predictions on large datasets periodically.

Using markers to isolate context segments.

A branch of ML using multi-layer neural networks to learn hierarchical representations, often excelling in vision, speech, and language.

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

Learning where data arrives sequentially and the model updates continuously, often under changing distributions.

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

A mismatch between training and deployment data distributions that can degrade model performance.

The relationship between inputs and outputs changes over time, requiring monitoring and model updates.

A structured collection of examples used to train/evaluate models; quality, bias, and coverage often dominate outcomes.

The internal space where learned representations live; operations here often correlate with semantics or generative factors.

Randomly zeroing activations during training to reduce co-adaptation and overfitting.

Converting text into discrete units (tokens) for modeling; subword tokenizers balance vocabulary size and coverage.

Networks with recurrent connections for sequences; largely supplanted by Transformers for many tasks.

Predicts masked tokens in a sequence, enabling bidirectional context; often used for embeddings rather than generation.

Maximum number of tokens the model can attend to in one forward pass; constrains long-document reasoning.

The text (and possibly other modalities) given to an LLM to condition its output behavior.

A high-priority instruction layer setting overarching behavior constraints for a chat model.

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