← Back to AI Factory Chat

AI Factory Glossary

145 technical terms and definitions

A B C D E F G H I J K L M N O P Q R S T U V W X Y Z All
Showing page 2 of 3 (145 entries)

refusal training, ai safety

Teach model when to refuse.

refused bequest, code ai

Subclass not using inherited methods.

regenerative thermal, environmental & sustainability

Regenerative thermal oxidizers recover heat through ceramic beds improving energy efficiency.

regex constraint, llm optimization

Regex constraints enforce pattern matching during text generation.

region-based captioning, multimodal ai

Caption specific image areas.

regret minimization,machine learning

Goal in online learning to minimize cumulative mistakes.

reinforcement graph gen, graph neural networks

Reinforcement learning guides graph generation through rewards for desired properties like drug-likeness.

reinforcement learning for nas, neural architecture

Use RL to search architectures.

reject option,ai safety

Allow model to say "I don't know".

relation extraction as pre-training, nlp

Use relation extraction as pretext task.

relation networks, neural architecture

Networks explicitly modeling pairwise relations.

relation-aware aggregation, graph neural networks

Relation-aware aggregation weights messages by edge types when updating node representations.

relational knowledge distillation, rkd, model compression

Transfer relationships between samples.

relative position bias, transformer

Learn relative position embeddings.

reliability-centered maintenance, rcm, production

Optimize maintenance strategy.

relm (regular expression language modeling),relm,regular expression language modeling,structured generation

Constrain generation using regex patterns.

renewable energy credits, environmental & sustainability

Renewable energy credits represent environmental attributes of renewable electricity generation.

renewable energy, environmental & sustainability

Renewable energy sources like solar and wind are increasingly adopted by fabs to reduce carbon footprint and energy costs.

renewal process, time series models

Renewal processes model event sequences where inter-arrival times are independent identically distributed random variables.

rényi differential privacy, training techniques

Rényi differential privacy generalizes epsilon-delta framework using Rényi divergence.

reorder point, supply chain & logistics

Reorder point is the inventory level triggering replenishment orders calculated from lead time demand and desired service level.

repetition penalty, llm optimization

Repetition penalty reduces probability of recently generated tokens.

replaced token detection, rtd, foundation model

Training objective in ELECTRA.

replanning, ai agents

Replanning revises strategies when plans fail or conditions change.

replica exchange, chemistry ai

Parallel simulations at different temperatures.

replicate,model hosting,simple

Replicate hosts ML models with simple API. Community models. Easy to deploy custom models.

repository understanding,code ai

Analyze entire codebases to understand architecture and dependencies.

representational similarity analysis, rsa, explainable ai

Compare representations across models.

representer point selection, explainable ai

Find most representative training examples.

request queuing, llm optimization

Request queuing manages incoming requests ensuring fair processing order.

residual post-norm, transformer

Normalization after residual addition.

residual stream analysis, explainable ai

Track information flow through residuals.

residual stress fa, failure analysis advanced

Residual stress analysis measures stress distributions in packages using techniques like Raman spectroscopy or warpage measurement.

resolution multiplier, model optimization

Resolution multipliers adjust input image size trading accuracy for speed.

response quality, training techniques

Response quality measures correctness helpfulness and safety of model outputs.

responsible ai principles,ethics

Guidelines for ethical AI development.

responsible ai,ethics,governance

Responsible AI encompasses fairness, transparency, safety, privacy. Governance frameworks guide practice.

responsible ai,rai,governance

Responsible AI: fairness, transparency, accountability, safety. Governance ensures ethical development and use.

retention mechanism, llm architecture

Retention mechanism computes attention-like aggregation through recursive formulation.

retinal image analysis,healthcare ai

Detect diseases from retinal images.

retnet, llm architecture

Retentive Network uses retention mechanism for parallel training and efficient inference.

retnet,llm architecture

Retention mechanism for efficient parallel training and recurrent inference.

retrieval-augmented language models, rag

Use retrieved documents to inform generation.

retro (retrieval-enhanced transformer),retro,retrieval-enhanced transformer,llm architecture

Architecture designed for retrieval augmentation.

retrosynthesis planning, chemistry ai

Plan synthetic routes backward from target.

retry logic, llm optimization

Retry logic attempts failed requests again with exponential backoff.

reverse osmosis, environmental & sustainability

Reverse osmosis purifies water by forcing it through semipermeable membranes removing dissolved solids.

reward hacking, ai safety

Reward hacking finds unintended ways to maximize reward without intended behavior.

reward model,preference,human

Reward model scores outputs based on human preferences. Train from comparisons. Used in RLHF.

reward model,preference,ranking

Reward model scores outputs by human preference. Trained on comparison data. Guides RL fine-tuning.