L
136 words and 470 other names, in italics, pointing to their entries.
- L1 loss → mean absolute error
- L1 norm → norm
- L1 regularization → regularization
- L1-penalized regression → lasso
- L2 distance → Euclidean distance
- L2 loss → mean squared error
- L2 norm → norm
- L2 regularization → regularization
- L2-regularized regression → ridge regression
- label
- label encoding → one-hot encoding
- label errors → label noise
- label noise
- label shift
- label smoothing
- labeled data → label
- labeling → data labeling
- labelled → label
- labelling → data labeling
- labelling of AI-generated content → AI-generated content labeling
- label-smoothing regularization → label smoothing
- labor displacement → technological unemployment
- LAD regression → quantile regression
- Lady Lovelace → Ada Lovelace
- Lady Lovelace's objection → Lovelace objection
- Lagrange multiplier
- Lagrangian → Lagrange multiplier
- Lagrangian duality → Lagrange multiplier
- LAION
- LAION e.V. → LAION
- LAION-2B-en → LAION-5B
- LAION-5B
- lakeFS → data versioning
- LALM → audio language model
- LambdaMART → learning to rank
- LambdaRank → learning to rank
- λ-return → eligibility trace
- LaMDA → LaMDA sentience controversy
- LaMDA sentience controversy
- Landau notation → big O notation
- landmark detection → keypoint detection
- lane centering → advanced driver-assistance system
- lane keeping assist → advanced driver-assistance system
- LangChain
- LangChain Expression Language → LangChain
- langchain-classic → LangChain
- Langevin dynamics → Markov chain Monte Carlo
- LangGraph
- LangGraph Platform → LangGraph
- LangGraph Studio → LangGraph
- langid → language identification
- LangSmith → LangChain
- language agent → AI agent
- Language and Machines → ALPAC report
- language detection → language identification
- language generation → natural language generation
- language ID → language identification
- language identification
- language model
- language modeling → language model
- language modeling objective → next-token prediction
- language modelling → language model
- language of thought → language of thought hypothesis
- language of thought hypothesis
- language technology → natural language processing
- language understanding → natural language understanding
- language-model head → unembedding
- large audio language model → audio language model
- large behavior model → robot foundation model
- large language model
- large language model grooming → LLM grooming
- large language model operations → LLMOps
- large language models → large language model
- large multimodal model → multimodal large language model
- large pretrained model → foundation model
- large reasoning model → reasoning model
- large vision-language model → vision-language model
- large-batch training → batch size
- large-margin classifier → margin
- large-model filing → algorithm filing
- Large-scale Artificial Intelligence Open Network → LAION
- large-scale training → distributed training
- laser radar → lidar
- laser scanner → lidar
- lasso
- lasso regression → lasso
- Last Dictionary → The Last Dictionary
- The Last Dictionary
- lastdictionary → The Last Dictionary
- lastdictionary.com → The Last Dictionary
- late chunking → chunking
- late fusion → sensor fusion
- late interaction
- latency
- latent → feature (interpretability)
- latent adversarial training
- latent code → latent space
- latent consistency model → consistency model
- latent diffusion → latent diffusion model
- latent diffusion model
- latent Dirichlet allocation
- latent factor model → matrix factorization
- latent reasoning
- latent representation → latent space
- latent semantic analysis
- latent semantic indexing → latent semantic analysis
- latent space
- latent steering → activation steering
- latent trait theory → item response theory
- latents → latent space
- latent-space reasoning → neuralese
- law of accelerating returns
- law of armed conflict → international humanitarian law
- law of averages → law of large numbers
- law of large numbers
- law of parsimony → Occam's razor
- LAWS → lethal autonomous weapons system
- Laws of Robotics → Three Laws of Robotics
- laws of war → international humanitarian law
- layer
- layer fusion → kernel fusion
- layer norm → layer normalization
- layer normalization
- layer skipping → early exit
- layered defense → defense in depth
- layered safeguards → defense in depth
- LayerNorm → layer normalization
- layer-wise relevance propagation
- layout analysis → document understanding
- lazy learning → k-nearest neighbors
- lazy training → neural tangent kernel
- L-BFGS → second-order optimization
- LBM → robot foundation model
- LCEL → LangChain
- LCM → consistency model
- LCM-LoRA → consistency model
- l-diversity → k-anonymity
- LDM → latent diffusion model
- LDP → local differential privacy
- Le Chat → Mistral AI
- LEACE
- lead agent → orchestrator–worker pattern
- leader board → leaderboard
- leaderboard
- league training → self-play
- leak → data leakage
- leakage → data leakage
- leaky integrate-and-fire → spiking neural network
- leaky rectified linear unit → leaky ReLU
- leaky ReLU
- Lean
- Lean 4 → Lean
- Lean mathematical library → Mathlib
- Lean proof assistant → Lean
- Lean prover → Lean
- Lean theorem prover → Lean
- learn → learning
- learnable parameter → parameter
- learned → learning
- learned optimization → mesa-optimization
- learned perceptual image patch similarity → perceptual loss
- learned representations → representation learning
- learned reward model → reward model
- learned robot control → robot learning
- learned sparse retrieval → sparse retrieval
- learning
- learning bias → inductive bias
- learning coefficient → singular learning theory
- learning curve
- learning from demonstration → imitation learning
- learning from human preferences → reinforcement learning from human feedback
- learning in context → in-context learning
- learning rate
- learning rate decay → learning rate schedule
- learning rate schedule
- learning rate warmup
- learning representations → representation learning
- learning set → training set
- learning task → task (machine learning)
- learning theory → statistical learning theory
- learning to learn → meta-learning
- learning to rank
- learning what to value → value learning
- learning with a teacher → supervised learning
- learning without a teacher → unsupervised learning
- learning-based robotics → robot learning
- learning-rate cooldown → warmup-stable-decay schedule
- learnt → learning
- least absolute shrinkage and selection operator → lasso
- least mean squares rule → delta rule
- least squares → ordinary least squares
- least-squares concept erasure → LEACE
- least-squares regression → linear regression
- least-to-most → least-to-most prompting
- least-to-most prompting
- leave-one-out baseline → REINFORCE leave-one-out
- leave-one-out cross-validation → cross-validation
- Yann LeCun
- left-to-right language model → causal language model
- legal personhood for AI → AI personhood
- Shane Legg
- legged locomotion → locomotion
- legged robot
- Legg–Hutter definition of intelligence → universal intelligence
- Legg–Hutter intelligence → universal intelligence
- legibility training → prover–verifier game
- Jan Leike
- lemma → stemming
- lemmatisation → stemming
- lemmatization → stemming
- Lemoine controversy → LaMDA sentience controversy
- Douglas Lenat
- LeNet
- LeNet-1 → LeNet
- LeNet-5 → LeNet
- length extrapolation → context extension
- length penalty → beam search
- length-controlled win rate → win rate
- Leslie G. Valiant → Leslie Valiant
- Leslie Gabriel Valiant → Leslie Valiant
- Less Wrong → LessWrong
- less-resourced language → low-resource language
- LessWrong
- lethal autonomous robotics → lethal autonomous weapons system
- lethal autonomous weapons system
- lethal trifecta
- lethal trifecta for AI agents → lethal trifecta
- Letta → MemGPT
- Level 2 → levels of driving automation
- Level 2+ → levels of driving automation
- Level 3 → levels of driving automation
- Level 4 → levels of driving automation
- Level 5 → levels of driving automation
- levels of agency → levels of autonomy
- levels of AGI
- Levels of AGI framework → levels of AGI
- levels of analysis → Marr's levels of analysis
- levels of automation → levels of autonomy
- levels of autonomy
- levels of driving automation
- Sergey Levine
- lexical disambiguation → word sense disambiguation
- lexical retrieval → sparse retrieval
- lexical search → sparse retrieval
- lexically constrained decoding → constrained decoding
- LfD → imitation learning
- Fei-Fei Li
- Li Fei-Fei → Fei-Fei Li
- liability for AI → AI liability
- Percy Liang
- Liang Wenfeng
- liar's dividend
- Libratus
- LibriLight → LibriSpeech
- LibriSpeech
- LibriTTS → LibriSpeech
- license-and-hire deal → reverse acquihire
- Lick → J. C. R. Licklider
- J. C. R. Licklider
- LID → language identification
- LIDA → cognitive architecture
- lidar
- lidar SLAM → simultaneous localization and mapping
- lie detection (AI) → AI lie detection
- LIF neuron → spiking neural network
- lifecycle hook → hook
- lifelong learning → continual learning
- light detection and ranging → lidar
- Light Gradient Boosting Machine → LightGBM
- LightGBM
- Lighthill report
- lights-out software development → dark factory
- likelihood
- likelihood function → likelihood
- likelihood-ratio gradient → policy gradient
- likeness rights → right of publicity
- LIME
- limited optimization → mild optimization
- limited rationality → bounded rationality
- limited risk → transparency obligations
- limited-risk AI → transparency obligations
- Linde–Buzo–Gray algorithm → vector quantization
- linear algebra
- linear artificial tomography → representation engineering
- linear attention
- linear CKA → centered kernel alignment
- linear classifier probe → linear probe
- linear concept erasure → concept erasure
- linear discriminant analysis
- linear function → linear transformation
- linear genetic programming → genetic programming
- linear layer → fully connected layer
- linear least squares → ordinary least squares
- linear map → linear transformation
- linear mapping → linear transformation
- linear mode connectivity → mode connectivity
- linear model → linear regression
- linear operator → linear transformation
- linear optimization → linear programming
- linear probe
- linear program → linear programming
- linear programming
- linear quadratic estimation → Kalman filter
- linear regression
- linear representation → linear representation hypothesis
- linear representation hypothesis
- linear separability
- linear temporal logic → temporal logic
- linear transformation
- linear transformer → linear attention
- linear warmup → learning rate warmup
- linear-chain CRF → conditional random field
- linearity hypothesis → linear representation hypothesis
- linearly separable → linear separability
- linear-quadratic regulator → optimal control
- link function → generalized linear model
- link prediction → knowledge graph embedding
- linkage → hierarchical clustering
- Linked Data → semantic web
- linked open data → semantic web
- Seppo Linnainmaa
- LinUCB → upper confidence bound
- Lion
- Lion optimizer → Lion
- Lipschitz → Lipschitz continuity
- Lipschitz condition → Lipschitz continuity
- Lipschitz constant → Lipschitz continuity
- Lipschitz continuity
- Lipschitz fairness → individual fairness
- liquid cooling
- liquid neural network
- liquid state machine → reservoir computing
- liquid time-constant network → liquid neural network
- LISA → Mila
- Lisp
- Lisp machine
- Lisp machines → Lisp machine
- LISt Processor → Lisp
- listwise ranking → learning to rank
- LiteRT
- LiteRT Micro → LiteRT
- live benchmark → contamination-resistant benchmark
- live facial recognition → remote biometric identification
- LiveBench → contamination-resistant benchmark
- LiveCodeBench → contamination-resistant benchmark
- Llama
- Llama 2 → Llama
- Llama 3 → Llama
- Llama 4 → Llama
- Llama Guard
- Llama Guard 2 → Llama Guard
- Llama Guard 3 → Llama Guard
- Llama Guard 4 → Llama Guard
- LlamaCloud → LlamaIndex
- llama.cpp
- LlamaGuard → Llama Guard
- LlamaHub → LlamaIndex
- LlamaIndex
- LlamaParse → LlamaIndex
- llama-server → llama.cpp
- LLC → local learning coefficient
- LLM → large language model
- LLM agent → AI agent
- LLM API → model API
- LLM API provider → inference provider
- LLM biology → model biology
- LLM cascade → model cascade
- LLM chain → prompt chaining
- LLM engineer → AI engineer
- LLM engineering → AI engineering
- LLM fingerprinting → model fingerprinting
- LLM gateway → AI gateway
- LLM grooming
- LLM guardrails → guardrails
- LLM injection → prompt injection
- LLM judge → LLM-as-a-judge
- LLM memory → agent memory
- LLM monitoring → LLM observability
- LLM nondeterminism → nondeterminism
- LLM observability
- LLM ops → LLMOps
- LLM optimization → generative engine optimization
- LLM orchestration → orchestration
- LLM proxy → AI gateway
- LLM psychology → machine psychology
- LLM response caching → semantic caching
- LLM router → model routing
- LLM routing → model routing
- LLM scaffolding → scaffolding
- LLM serving → model serving
- LLM stack → AI stack
- LLM steering → steering
- LLM tracing → LLM observability
- LLM understanding → machine understanding
- LLM unlearning → machine unlearning
- LLM watermarking → AI watermarking
- LLM workflow → agentic workflow
- LLM wrapper → wrapper
- LLM-as-a-judge
- LLM-as-judge → LLM-as-a-judge
- LLMflation → inference cost
- LLM.int8()
- LLM-ism → LLM-isms
- LLM-isms
- LLMO → generative engine optimization
- LLMOps
- LLMs → large language model
- llms-full.txt → llms.txt
- llms.txt
- LLN → law of large numbers
- Lloyd–Forgy algorithm → k-means clustering
- Lloyd's algorithm → k-means clustering
- LM → language model
- LM head → unembedding
- LMArena
- lmarena.ai → LMArena
- LM-based red teaming → automated red teaming
- lm-evaluation-harness → evaluation harness
- LMM → multimodal large language model
- LMS algorithm → delta rule
- LMSYS Chatbot Arena → LMArena
- LN → layer normalization
- LOAC → international humanitarian law
- load-balancing loss → router
- LOAR → law of accelerating returns
- lobotomised → lobotomized
- lobotomize → lobotomized
- lobotomized
- lobotomy → lobotomized
- local AI → on-device AI
- local attention → sliding window attention
- local beam search → beam search
- local differential privacy
- local DP → local differential privacy
- local explanation
- local inference → on-device AI
- local interpretability → local explanation
- Local Interpretable Model-agnostic Explanations → LIME
- Local Law 144 → automated employment decision tool
- local learning coefficient
- local LLM → on-device AI
- local maximum → local minimum
- local minima → local minimum
- local minimum
- local optimum → local minimum
- local outlier factor → anomaly detection
- local planner → path planning
- local recurrence theory → recurrent processing theory
- local replacement model → replacement model
- local search
- local search algorithm → local search
- local SGD
- local stochastic gradient descent → local SGD
- local surrogate → surrogate model
- local–global attention → sliding window attention
- locality-sensitive hashing
- lock-in → value lock-in
- loco-manipulation → locomotion
- locomotion
- LOD → semantic web
- Loebner Prize
- log loss
- log odds → logit
- log probability
- log score → log loss
- logarithmic loss → log loss
- logarithmic score → proper scoring rule
- logarithmic scoring rule → log loss
- log-density → probability density function
- logic program → logic programming
- logic programming
- Logic Theorist
- Logic Theory Machine → Logic Theorist
- logical behaviorism → behaviorism
- logical consequence → deductive reasoning
- logical induction
- logical induction criterion → logical induction
- logical inductor → logical induction
- logical omniscience → logical uncertainty
- logical uncertainty
- logicist AI → symbolic AI
- logistic function → sigmoid function
- logistic loss → log loss
- logistic regression
- logistic sigmoid → sigmoid function
- logit
- logit attribution → direct logit attribution
- logit bias
- logit difference attribution → direct logit attribution
- logit function → logit
- logit lens
- logit model → logistic regression
- logit regression → logistic regression
- logits → logit
- log-likelihood
- log-likelihood ratio → log-likelihood
- log-mel spectrogram → mel spectrogram
- log-odds → logit
- log-prob → log probability
- logprob → log probability
- logprobs → log probability
- log-sum-exp trick → numerical stability
- log-Z penalty → z-loss
- long context
- long short-term memory
- long tail
- long timelines → AI timelines
- long-context LLM → long context
- long-context model → long context
- long-context modeling → long context
- Longformer → sliding window attention
- long-horizon → long-horizon task
- long-horizon agent → long-horizon task
- long-horizon planning → long-horizon task
- long-horizon task
- long-running task → long-horizon task
- long-tail knowledge → long tail
- long-tailed distribution → long tail
- long-term memory
- longtermism
- longtermist → longtermism
- LOOCV → cross-validation
- look-ahead bias → data leakage
- look-ahead mask → causal mask
- lookup-table argument → Blockhead
- loop closure → simultaneous localization and mapping
- loopy belief propagation → belief propagation
- LoRA → low-rank adaptation
- LoRA adapter → low-rank adaptation
- LoReFT → representation fine-tuning
- losing control of AI → loss of control
- loss → loss function
- loss blow-up → loss spike
- loss curve
- loss divergence → training instability
- loss floor → irreducible loss
- loss function
- loss landscape
- loss of control
- loss of plasticity
- loss of trainability → loss of plasticity
- loss plot → loss curve
- loss scaling → mixed-precision training
- loss spike
- loss spikes → loss spike
- loss surface → loss landscape
- loss-of-control risk → loss of control
- lossy compression of the web → blurry JPEG of the web
- lost in the middle
- lost-in-the-middle effect → lost in the middle
- LOT → language of thought hypothesis
- Lotfi A. Zadeh → Lotfi Zadeh
- Lotfi Aliasker Zadeh → Lotfi Zadeh
- LOTH → language of thought hypothesis
- lottery ticket hypothesis
- lottery tickets → lottery ticket hypothesis
- Ada Lovelace
- Lovelace objection
- Lovelace test → Lovelace objection
- low-bit quantization → quantization
- low-impact AI → impact measure
- low-loss path → mode connectivity
- low-rank adaptation
- low-rank adapters → low-rank adaptation
- low-rank approximation
- low-rank decomposition → low-rank approximation
- low-rank factorization → low-rank approximation
- low-resource language
- low-resource languages → low-resource language
- low-resource NLP → low-resource language
- low-shot learning → few-shot learning
- LP → linear programming
- LPDDR → DRAM
- LPIPS → perceptual loss
- LQR → optimal control
- LR → learning rate
- LR schedule → learning rate schedule
- LR warmup → learning rate warmup
- LReLU → leaky ReLU
- LRH → linear representation hypothesis
- LRL → low-resource language
- LRM → reasoning model
- LRP → layer-wise relevance propagation
- LSA → latent semantic analysis
- LSH → locality-sensitive hashing
- LSI → latent semantic analysis
- LSM → reservoir computing
- L-smoothness → Lipschitz continuity
- LSR → label smoothing
- LSTM → long short-term memory
- LSTM network → long short-term memory
- LT → Logic Theorist
- LTC network → liquid neural network
- LTH → lottery ticket hypothesis
- LTL → temporal logic
- LTM → long-term memory
- LTR → learning to rank
- Lucas–Kanade method → optical flow
- Lucas–Penrose argument
- Luddism → Luddite
- Luddite
- Luddite fallacy → Luddite
- lump of labor fallacy → technological unemployment
- lurking variable → confounding variable
- LVLM → vision-language model
- LW → LessWrong