AI Factory Glossary

prompt tuning, prompting techniques

Prompt tuning learns continuous prompt embeddings through gradient descent.

prompt tuning,fine-tuning

Learn soft prompts (continuous vectors) while keeping the model frozen.

prompt versioning,version control

Version control prompts like code. Track changes, A/B test versions, rollback if quality drops.

prompt weighting, generative models

Emphasize prompt parts differently.

prompt weighting,prompt engineering

Assign different importance to different parts of the prompt.

prompt-based continual learning, continual learning

Use prompts to separate tasks.

prompt-to-prompt editing,generative models

Edit images by modifying text prompts.

prompt-to-prompt, multimodal ai

Prompt-to-prompt editing modifies images by adjusting text prompts while preserving structure.

prompt,prompting,instruction

Prompt = text you feed the model. Clear role, task, input format, and constraints = more reliable and controllable outputs.

promptable segmentation,computer vision

Segment using various prompt types.

promptfoo,testing,eval

Promptfoo tests prompts and models. CLI tool. Compare outputs.

promptlayer,logging,versioning

PromptLayer logs and versions prompts. Track changes, A/B test.

pronoun resolution, nlp

Resolve pronouns to antecedents.

proof generation,reasoning

Create mathematical proofs.

propensity score rec, recommendation systems

Propensity scores weight training examples by inverse probability of exposure to debias recommendation models.

property inference, privacy

Infer properties of training data.

property-based test generation, code ai

Generate property tests.

property-based testing,software testing

Test general properties rather than examples.

prophet, time series models

Prophet is an additive time series model decomposing signals into trend seasonality and holiday components with automatic changepoint detection.

proportional task sampling, multi-task learning

Sample by dataset size.

proposal,business,write

AI drafts business proposals. Structure, persuasion.

proposition retrieval,rag

Retrieve atomic facts/propositions instead of passages.

proprietary model, llm architecture

Proprietary models have restricted access to weights and implementation.

protected health information detection, phi, healthcare ai

Identify sensitive medical data.

protective capacity, manufacturing operations

Protective capacity maintains deliberate excess at non-constraints enabling system flexibility.

protein design,healthcare ai

Design proteins with desired properties.

protein function prediction from text, healthcare ai

Infer protein function from descriptions.

protein structure prediction,healthcare ai

Predict 3D protein structures (AlphaFold).

protein-ligand binding, healthcare ai

Interaction between protein and small molecule.

protobuf,binary,grpc

Protocol Buffers is binary serialization. Efficient, typed. Used with gRPC.

prototype learning, explainable ai

Learn representative prototypes for each class.

prototype testing, design

Test early versions.

prototypical contrastive learning, self-supervised learning

Contrast against learned prototypes.

prototypical networks,few-shot learning

Learn embeddings where examples from same class cluster together.

provenance tracking, rag

Track information sources.

provenance tracking, security

Track model lineage and modifications.

provenance tracking,trust & safety

Record origin and modifications of content.

proximal policy optimization, ppo, reinforcement learning

Clipped objective for stable updates.

proximity effect, signal & power integrity

Proximity effect alters current distribution in adjacent conductors increasing loss and inductance.

proximity gettering, process

Getter near front surface.

proxylessnas, neural architecture

Search directly on target hardware.

proxylessnas, neural architecture search

ProxylessNAS directly learns architectures on target hardware by training over-parameterized networks with path-level binarization and latency constraints.

pruning gaussians, 3d vision

Remove low-opacity Gaussians.

pruning, model optimization

Pruning removes unnecessary network components reducing size and computation.

pruning,model optimization

Remove weights or neurons that contribute little to performance.

pruning,sparsity,compression

Pruning removes unnecessary weights. Structured pruning removes neurons/heads. Reduces model size and compute.

pseudo relevance, rag

Pseudo relevance feedback expands queries using top retrieved results.

pseudo-count methods, reinforcement learning

Estimate state visitation counts.

pseudo-labeling with confidence, semi-supervised learning

Use high-confidence predictions as labels.

pseudo-labeling, advanced training

Pseudo-labeling assigns predicted labels to unlabeled data treating them as ground truth for semi-supervised learning.