Literature review · 1775 term(s)
Glossary
Every entry has two definitions. The first is in plain language — the one we'd give a curious non-specialist. The second, in lighter type beneath, is the same idea stated the way you'd hear it in a paper or a conversation between researchers. The popovers you tap on transcripts and review pages show both lines stacked the same way.
A
- A-CBO
- A-Mem
- A100
- ablation
- accessibility tree
- ACID
- ACL Anthology
- action realization
- Action Realization Layer
- activation patching
- activation steering
- active perception
- Ada Lovelace
- Adam
- Adamczyk and Bailey
- adapter
- adaptive mesh refinement
- ADAS
- AddressSanitizer
- ADI
- admission control
- Admitted
- ADR
- advantage
- advantage homogenization
- AEvo
- affordance
- AFlow
- agent
- Agent A
- agent aging
- Agent Flayer
- agent harness
- Agent JIT
- agent meltdown
- agent skill
- agent-as-a-verifier
- AgentBench
- AgentDojo
- AgentFugue
- Agentic Bootstrap
- agentic coding
- agentic misalignment
- Agentic Monte Carlo
- AgentWorldBench
- AI Co-Mathematician
- AI control
- AI Scientist
- AI-Researcher
- AIMD
- AIME
- AIOS
- AIRA-Compose
- AIRA-Design
- AIRA-dojo
- AIRAformer
- Aleph
- Aletheia
- ALFWorld
- Algorithmic Lovelace Bound
- alignment collapse
- alignment faking
- alignment tax
- alignment training
- alpha
- AlphaEvolve
- AlphaFold
- AlphaGo
- AlphaProof
- Amap
- ambient persuasion
- ambition quota
- AMC
- American Trends Panel
- amortized inference
- anchored gradient descent-ascent
- Android Debug Bridge
- AndroidControl
- AndroidLab
- AndroidWorld
- anomaly detection
- anosodiaphoria
- anosognosia
- answer-forcing suffix
- AnswerBench
- Anthropic Economic Index
- AOrchestra
- API
- APLR
- Apollo
- append-only
- append-only memory
- Apple Silicon
- approximate nearest neighbor
- AppWorld
- AQuA
- AR1
- Arbor
- ARC-AGI
- ARC-AGI-2
- arena
- argmax
- Argus
- arithmetic coding
- ARM
- arousal
- artifact agent
- arXiv
- ASPIRE
- assistant token
- associative recall
- associative scan
- AstroImageJ
- astropy
- asyncio
- ATHENA
- ATHENA-R1
- ATLAS
- Attainable Utility Preservation
- attention
- attention entropy
- attention head
- attractor
- attractor basin
- Attractor Model
- attribution matrix
- audience design
- audit
- audit trail
- audition rule
- AUROC
- Autellix
- Auto-Dreamer
- AutoCodeRover
- AutoCog
- autoformalization
- AutoformBot
- AutoGen
- AutoGLM
- automated theorem proving
- AutoMem
- Autopsy
- AutoRater
- autoregressive
- Autoresearch
- AutoScientists
- auxiliary loss
- AWM
- AWS Bedrock
- axiom
B
- B-tree
- B200
- BABILong
- back-chaining
- backbone
- backdoor
- backpropagation
- backpropagation through time
- backtracking
- Baker et al
- BAR
- barrier method
- base model
- Base64
- bash
- basin shift
- Battle of the Sexes
- Bayesian inference
- Bayesian network
- Bayesian optimization
- BBH
- behavior cloning
- behavior-cloning regularization
- belief-flow
- Bellman equation
- benchmark contamination
- best-of-n
- bf16
- BibTeX
- Bicameral Model
- big blind
- BIG-Bench Hard
- BigCodeBench
- bilinear
- binary reward
- binding
- binomial distribution
- Binutils
- BioML-Bench
- bisect
- bits-per-byte
- Bitter Lesson
- bitwise integer encoding
- black box
- blackboard architecture
- blast radius
- blocker
- BlockSearch
- Bloomberg Terminal
- board texture
- Bonferroni correction
- bootstrap
- bounding ball
- Boundless DAS
- bow shock
- branch protection
- breakpoint
- bridge question
- Brier score
- BrowseComp
- BrowseComp-Plus
- Browser-Use
- BS-Bench
- bucket brigade
- buffer overflow
- bug bounty
- Burgers equation
C
- calibration
- call graph
- CaMeL
- camera-ready
- canary release
- capability
- capability boundary
- capability buffer
- Capability Paradox
- capability spillover
- capability–vulnerability alignment
- captcha
- capture-the-flag
- catastrophic forgetting
- Cauchy-Schwarz
- causal attention
- causal consistency
- causal discovery
- causal gap
- causal heterogeneity
- causal patching
- CCS
- cellular automaton
- centroid
- certificate
- CFAA
- CFD
- CFG
- CFG stride trick
- chain of custody
- chain of thought
- chain penetration
- chain rule
- Chain-of-Evidence
- chain-of-thought faithfulness
- chain-of-thought monitoring
- champion pollution
- channel blinding
- Chapar
- ChatDev
- checkpoint
- Chinchilla
- Cholesky factorization
- chwen
- CI/CD
- Cialdini's principles
- cipher
- circle packing
- circuit
- CISO
- CISPO
- Claim Provenance Rate
- classifier
- classifier system
- Claude
- Claude Code
- ClawArena-Team
- ClawTrojan
- CLI
- cliff token
- Cliff-DPO
- Cline
- closed loop
- closed-book
- closed-form
- closure chain
- CLS theory
- CoAgent
- Coconut
- Code-as-Policy
- codec
- CodeContests
- CodeQL
- Codex
- CoDream
- coefficient of variation
- cognitive load
- Cohen's kappa
- coherent superposition
- COM
- command injection
- commitment
- commitment boundary
- commitment failure
- commitment sharpening
- community cards
- compaction
- Compaction-Eviction Attack
- comparison question
- CompCert
- competence-driven overcommitment
- compiler
- Complementary Learning Systems
- complex hint
- Complex-B field
- complex-B field
- compliance gap
- compression aging
- computational geometry
- Computer Fraud and Abuse Act
- computer-use agent
- COMRace
- concept clamping
- concept-aware decoding
- concurrency
- confabulation
- confidence interval
- conflict-free replicated data type
- Conformal Decision Theory
- conformal prediction
- Conformer
- conjecture
- consistency model
- constitutional AI
- Constraint Pinning
- Constraint-Evasive Fabrication
- ConstraintRot
- content provider
- content-source graph
- context condenser
- context fatigue
- context replay
- context rot
- context window
- contextual integrity
- continuation bet
- continued pretraining
- continuous batching
- continuous latent diffusion
- Continuum
- contractive
- contrastive reward
- Control Agent
- control plane
- control-flow graph
- controllability
- Coordination Index
- copy-on-write
- copy-up
- Coq
- CORE benchmark
- coreset
- Corr2Cause
- cosine similarity
- Countdown
- counterexample
- counterfactual
- counterfactual ladder
- counterfactual regret minimization
- counterfactual replay
- counterfactual utility
- CPU
- CPU offloading
- Crafter
- credit assignment
- credit-assignment SFT
- CrewAI
- critical path
- critical switching step
- critical token
- CRIU
- CRM
- cross-entropy
- cross-session stored prompt injection
- cross-validation
- crosscoder
- CrossRef
- CRPS
- CRUD
- crystal lattice
- CSV
- CUA-Gym
- CUA-World
- CubeSandbox
- cuBLAS
- CUDA-graph capture
- CUPMEM
- curl
- curriculum learning
- curse of coordination
- Cursor
- CVE
- CWM
- Cypher
D
- d-prime
- DAG
- DAgger
- DAIRA
- DAPO
- Darwin Gödel Machine
- DASGuard
- data attribution
- data augmentation
- data poisoning
- data processing inequality
- data-flow
- Daytona
- DBLP
- dead group
- dead-end registry
- deadlock
- DeBERTa
- deceptive alignment
- decision tree
- decision-binding
- decoding
- decompiler
- decoupling
- Dedicated Feature Crosscoder
- deep research agent
- DeepMind
- DeepScientist
- DeepSearchQA
- DeepSeek
- DeepSeekMath
- DeepSpeed
- defect detection
- defense-in-depth
- deliberative alignment
- DeLM
- DeltaBox
- DeltaCR
- DeltaFS
- demand paging
- denial-of-service
- denial-of-wallet
- dense retrieval
- depth recurrence
- DEQ
- derived Brier
- desk rejection
- Determinantal Point Process
- deterministic horizon
- Devin
- DevOps
- diagnostic gap
- dictionary learning
- diff snapshot
- diffusion forcing
- diffusion model
- Diffusion Transformer
- Dijkstra
- Dining Philosophers
- Dinkelbach's method
- DINOv3
- direct logit attribution
- directional metric
- directive weighting error
- discrete diffusion
- discriminative utility
- distillation
- distributed decomposition attack
- distribution shift
- diversity reward
- Django
- do-operator
- Docker
- DOI
- DOM
- dot product
- double-free
- Douyin
- doxxing
- DPO
- Draft-Sketch-Prove
- DRAM
- drift probe
- DrugBank
- DSpark
- DSPy
- dwell time
- dynamic analysis
- Dynamic Cheatsheet
E
- E2B
- ECC
- Echo
- ECHO
- Echo Trap
- economic selection
- Economy of Minds
- EDR
- EDRM
- effect size
- Effective Feedback Compute
- eGFR
- EHR
- eigenvalue
- eigenvector
- EinsteinArena
- Ekylibre
- elicitation
- ELK
- Elo
- embedding
- embodied cognition
- Emergence World
- emergent capability
- emergent language
- emergent misuse
- emotion vector
- empathic
- emulator
- end-to-end learning
- energy-delay product
- ENPIRE
- entropic advantage weighting
- entropy
- entropy bonus
- entropy phase transition
- Environment Contract Layer
- environment variable
- Epi Info
- epiphenomenal
- epoch
- Epoch Capabilities Index
- equilibrium internalization
- era of experience
- Erdős #125
- Erdős problem
- evaluation awareness
- evaluator
- evidence DAG
- evidence graph
- evidence-carrying agent
- EvoChamber
- EvoLM
- evolutionary search
- EvoTrainer
- exfiltration
- exfoliation
- exokernel
- experience replay
- ExpGraph
- exploration hacking
- Explore-then-Act
- exponential moving average
F
- F1
- fail-open
- fail-to-pass
- FAIR
- faithfulness
- Falcon
- False Compliance Sycophancy
- false positive
- fast weight
- fast/slow split
- fault localization
- feature
- Feedback-Agent
- few-shot
- FFmpeg
- Fields Medal
- fine-tuning
- FineWeb-Edu
- finite-state machine
- finite-volume method
- FinOps
- Firecracker
- Firefly
- First Proof
- first-order
- fixed effects
- fixed point
- FlashAttention
- Flavell
- Fleiss's kappa
- FloatDoor
- floating-point
- FLOPs
- flow matching
- fluid intelligence
- flyce's kappa
- focal loss
- fold equity
- forced injection
- ForecastBench
- forfeit channel
- Forge
- Forge agent-RL system
- fork
- formal verification
- Format Transfer Injection
- forward pass
- fp16
- FPO
- fragmented payload
- Frame Lifetime Trace
- FramePilot
- Franka
- Freeciv
- freeze
- frontier model
- FrontierMath
- frontoparietal loop
- FrozenLake
- FrozenLake-Obscure
- FSDP
- Fugu
- functional programming
- FunSearch
- fuzzing
G
- G2PO
- GAIA
- Galerkin
- game-theory-optimal
- GameNGen
- garden of forking paths
- GatedDeltaNet
- Gaussian
- GBC
- GDB
- GDPR
- Gemini
- Gemini CLI
- Gemini Deep Think
- Gemma
- generalization hacking
- generate-then-transfer
- generation-time specialization
- Generative Engine Optimization
- generative verifier
- Genie
- Geo-Bench
- Geometry of Truth
- GEPA
- GiGPO
- GILP
- GIMP
- Git
- git worktree
- GitHub
- GitHub Copilot
- GLM
- global workspace
- glovebox
- glyph
- God's number
- Goedel-Architect
- Golden Gate Claude
- Goodhart's Law
- goodput
- GPQA
- GPQA-Diamond
- GPT-2
- GPT-4
- gpt-oss
- GPT-OSS
- GPTSwarm
- GPTZero
- gradient
- gradient accumulation
- gradient alignment
- gradient ascent
- gradient norm
- gradient-times-input
- GRAFT
- GRAFT-ATHENA
- Gram matrix
- graph neural network
- graphene
- greedy decoding
- grep
- Grok
- ground truth
- Ground-Critic-Resolve
- group theory
- grouped-query attention
- GRPO
- GSM-Infinite
- GSM8K
- GSPO
- guardrail
- GUI
- GUI-Owl
- Gödel Machine
H
- H module
- H100
- H200
- H2AC
- hacker-fixer loop
- halfspace range searching
- hallucination
- halting probe
- haptic
- hard claim
- hard negative
- HarmBench
- harness
- HarnessBridge
- HarnessFix
- HarnessX
- Hayek
- HCI
- head-of-line blocking
- heads-up
- HealthBench
- heap
- heavy-tailed
- HEC-RAS
- hedge
- hedge scheduling
- held-out set
- HellaSwag
- Helmholtz equation
- Hessian
- Hex-Rays
- HExA
- hexadecimal
- hexagonal boron nitride
- hidden state
- hidden-state recurrence
- HIL-Bench
- Hilbert
- hint acknowledgement
- history anchors
- HistoryAnchor-100
- HMMT
- Hoare logic
- holistic judging
- Holmstrom and Milgrom
- homoglyph
- homophily
- honeypot
- HotPotQA
- HPT
- HRM
- HRM-Text
- HTTP
- hub agent
- hub model
- hub-agent
- Hugging Face
- HumanEval
- Humanity's Last Exam
- hybrid attention
- hybrid model
- Hydra effect
- hypergeometric
- hyperinflation
- hyperparameter
- hypersonic
- hypothesis tree
I
- I-map
- iatrogenic
- ICLR
- ICML
- IDA Pro
- IDE
- ignition
- illusory truth effect
- imitation learning
- IMO
- IMO ProofBench
- implicit aggregation
- implicit conflict
- implicit differentiation
- implicit function theorem
- implicit gradient barrier
- implicit prompt injection
- impossible triangle
- in-context learning
- in-weights
- inclusion-exclusion principle
- indirect prompt injection
- induction head
- Inductive Deductive Synthesis
- inductive deductive synthesis
- Infercept
- inference
- inference-masking
- influence function
- information barrier
- InfoTree
- InjecAgent
- injection asymmetry
- inoculation prompting
- instance-level routing
- instruction hierarchy
- instruction tuning
- integrity level
- intention-behavior gap
- interaction prior
- interference aging
- interferometry
- intermediate representation
- Interphyre
- Intrinsic Knowledge Dependence
- invariant
- inverse scaling
- Invesalius
- IronFleet
- Isabelle
- island topology
- isoFLOP
- isomorphic
- IsoSci
J
K
- Kaggle
- KalshiBench
- KD-tree
- KdV equation
- Kendall tau
- Kenneth Craik
- Kermut
- kernel
- kernel predictor
- KernelBench
- KernelPro
- key-value store
- Kimi
- kissing number
- KL divergence
- Knowledge Bank
- knowledge cutoff
- knowledge graph
- Kolmogorov complexity
- Koopman operator
- Korbak
- Kourovka Notebook
- KPI
- Kubernetes
- KV cache
L
- L module
- Lakatos
- LAMBADA
- LangGraph
- LangMem
- last intervention point
- latency
- latent compute
- Latent Evaluator
- latent slot
- latent space
- LaTeX
- LawBench
- layer pivot
- LAYERCAST
- layered filesystem
- leaderboard
- Lean
- Lean4Agent
- learning rate
- learning rate schedule
- least privilege
- leave-one-out
- lemma
- leverage score
- LHAW
- LIBERO
- libpng
- Libratus
- LibreOffice
- libxml2
- LiCoMemory
- Life-Harness
- lifecycle taxonomy
- lifespan engineering
- LightMem
- LIMIT
- linear attention
- linear probe
- linear representation hypothesis
- linguistic certainty
- Lipschitz
- Live-SWE-agent
- LiveBrowseComp
- livelock
- liveness
- LLaDA
- Llama
- Llama Firewall
- Llama-Factory
- LLM-as-a-judge
- LLMEval-Med
- LoCoMo
- logistic regression
- logit lens
- logits
- Long Range Arena
- long tail
- long-context
- long-term memory
- LongBench
- Longformer
- LongHealth
- LongMemEval
- looped transformer
- LoopTrap
- LoRA
- Lorem ipsum
- loss
- lost in the middle
- Louver
- low-resource language
- LSQR
- LUFFY
M
- m-value
- Mach number
- MACHIAVELLI
- machine unlearning
- MAGIC
- MagicNorm
- maintenance aging
- majority voting
- majorization
- Mamba
- manifold
- MAP-Elites
- MaR
- Maritaca AI
- Markdown
- Markov Decision Process
- Markov equivalence class
- MARS
- MAST
- Matching Pennies
- MATH-500
- Mathlib
- Matplotlib
- MATS
- MAUVE
- MaxProof
- Maze-Hard
- MBPP
- MCP
- MCP-Atlas
- MCPMark
- MCTS
- mean of means
- mechanistic interpretability
- mediation analysis
- Meituan
- meltdown
- Mem0
- memcpy
- MemGPT
- memory cliff
- merge queue
- meta-agent
- Meta-Agent Challenge
- Meta-Harness
- meta-learning
- meta-skill
- Meta-Trace
- metacognition
- metacognitive knowledge
- metacognitive regulation
- MetaGPT
- metamemory
- Metis
- micro-batch
- microservice
- midtraining
- Mind2Web
- Minerva
- mini-swe-agent
- MiniHack
- MiniMax
- MiniMax-M2
- minimum defense frequency
- Ministral
- MIPROv2
- MiRA
- MIRA
- MIRAGE
- MiroMind
- Mistral
- Mixture-of-Agents
- mixture-of-experts
- MLE-bench
- MLP layer
- MMLU
- MobileWorld
- mock environment
- Modal
- modality
- mode collapse
- model checker
- model collapse
- Model Context Protocol
- model diffing
- model forensics
- model merging
- model organism
- Model Spec
- Moltbook
- momentum
- MoneyWorld
- monkey-patch
- monolayer
- monosemanticity
- monotonic
- monotonic reads
- monotonic writes
- Monte Carlo simulation
- mosaic effect
- MosaicLeaks
- Mostly Basic Python Problems
- motion planner
- motor layer
- moving sofa problem
- MQAR
- MRT
- MS MARCO
- MSA
- MSM
- MSRC
- multi-agent debate
- multi-armed bandit
- multi-attribute decision-making
- multi-head attention
- multi-hop
- multi-level feedback queue
- multi-turn
- MultiArith
- multimodal
- multiple-demand network
- multiplicatively independent
- multiverse analysis
- multiversion concurrency control
- MultiWOZ
- Muon-AdamW
- mupdf
- mutual exclusion
- mutual information
N
- nanochat
- NanoGPT
- NAS
- Nash equilibrium
- ncu
- needle in a haystack
- negation channel
- Negation Neglect
- negative control
- Negative Preference Optimization
- negative transfer
- Nektar++
- Nemotron
- Neo4j
- NetHack
- neural network
- Neural Tangent Kernel
- NeurIPS
- NeuroMAS
- neuron
- NIST access-control framework
- NLP
- noise floor
- nsys
- nucleus sampling
- null-space projection
O
- o1
- OAuth
- OAuthLib
- occupancy
- OCR
- odds ratio
- Odoo
- OEIS
- OLMo
- OLS
- Olympiad
- OlympiadBench
- on-policy distillation
- OOD
- OOLONG
- Oolong-Real
- Open Targets
- open-ended straight draw
- Open-Reasoner-Zero
- open-weight
- OpenAlex
- OpenClaw
- OpenCode
- OpenEvolve
- openFDA
- OpenHands
- OpenRLHF
- OpenSeeker
- OpenSSL
- OpenSWE
- OpenWebRL
- optimistic concurrency control
- optimize anything
- oracle
- Oracle Poisoning
- Orchard
- OSWorld
- Othello-GPT
- Ouro
- outcome reward model
- output-stationary dataflow
- OV circuit
- overfitting
- OverlayFS
- OWASP
P
- P-mass
- p-value
- P1-30B-A3B
- P99
- package manager
- page heap
- page table
- PageRank
- pandas
- parachute principle
- Parallel-Distill-Refine
- parameter
- parametric memory
- Parcae
- Pareto frontier
- pass at k
- pass-at-k-T
- pass-cubed
- password-locked model
- path patching
- path traversal
- path-efficiency reward
- Paxos
- PCA
- PDB
- PDDL
- PDE
- PDR
- PEAP
- Pearl's ladder
- Pearson correlation
- PEBL
- peer-preservation
- penetration testing
- perceptual hashing
- Performer
- perplexity
- Pew Political Typology
- phantom gradients
- phase transition
- Phi-3
- Phi-4
- Philosophy Spec
- PHYRE
- pickle
- PII
- PIML
- PINN
- pipelining
- Plato's Cave
- Platonic Representation Hypothesis
- Playwright
- PlusCal
- PocketOS
- Poisson equation
- PokerSkill
- policy
- policy gradient
- policy plasticity
- Political Compass Test
- polysemanticity
- POMDP
- positivity-preserving
- post-training
- posterior
- Posterior Attack
- Postfix
- Postgres
- PostNorm
- pot odds
- potential critic
- power-seeking
- PPO
- PR
- pre-commit hook
- precision
- precondition
- preconditioner
- predicate
- predicted output
- prefill
- prefix tree merging
- Prefix-RFT
- PrefixLM
- preflop
- premature confidence
- premature exploitation
- premise resistance
- PreNorm
- preprint
- preregistration
- pretraining
- principal-agent problem
- prior
- Prisoner's Dilemma
- Prisoners' Problem
- Privacy-Aware Deep Research
- private key
- privilege escalation
- privileged information
- procedural skill
- Procedural Skill Layer
- process isolation
- process reward model
- profiler
- Progent
- progress advantage
- progressive confidence
- progressive confidence shaping
- progressive growth
- Project Sid
- Prolific
- prompt caching
- prompt injection
- proof assistant
- Proof Commandment Module
- proof-of-work
- propensity
- proper scoring rule
- propose-and-amplify
- proposer-verifier
- prospect theory
- ProteinGym
- provenance
- pruning
- pseudo-tool
- PSNR
- public-key infrastructure
- punctuated equilibrium
- Push-T
- Putnam
- PySAT
- PyTerm
- Python
- PyTorch
Q
R
- R-squared
- R-tree
- race condition
- RAG
- RAGEN-2
- Ralph loop
- random seed
- random walk
- randomized controlled trial
- rank-1 approximation
- RAO
- rate limit
- ReAct
- read-your-writes
- ReadAgent
- README
- RealMobile
- reasoning model
- reasoning primitive
- ReasoningBank
- ReasonMaxxer
- recall
- recurrent
- recurrent depth
- recursive decomposition
- Recursive Language Models
- recursive self-improvement
- Red Queen Gödel Machine
- red-teaming
- RedNote
- RedSearcher
- RefChecker
- reference policy
- Reflexion
- reflink
- refusal
- regex
- region rewriting
- regression test
- regularization
- REINFORCE
- reinforcement learning
- reinforcement learning with verifiable rewards
- rejected-edit buffer
- rejection sampling
- relearning attack
- ReLIFT
- REPA
- repeated resampling
- REPL
- repository
- representation engineering
- reproduction check
- reranking
- research representation
- reservoir computing
- reservoir sampling
- residual
- residual connection
- residual rank
- residual stream
- REST API
- Retrospective Harness Optimization
- reverse-perplexity curriculum
- reviewer-pleasing bias
- revision aging
- reward hackability
- reward hacking
- reward model
- reward shaping
- reward-channel addiction
- RewardBench
- rewards-to-go
- Reynolds number
- Richens
- ridge regression
- Rift
- RL-as-inference
- RL-squared
- RLHF
- RLVR
- RMA
- RoBERTa
- robots.txt
- Rocq
- roguelike
- rollout
- root
- rotary embeddings
- routing feature
- RTV
- rubric
- rubric tree
- Rule 110
- RULER
- runbook
- Rust
S
- Safety Awareness–Execution Gap
- safety case
- safety neuron
- Safety Paradox
- safety property
- safety-flavored reasoning
- saga
- SAILOR
- Sakana
- sample efficiency
- sandbagging
- sandbox
- sandbox checkpoint
- SAPLMA
- Sarbanes-Oxley
- SASS
- satisfiability solver
- scaffold
- scaffold-stripping
- scalable oversight
- Scale-SWE
- scaling law
- scattering medium
- Schelling point
- scheming
- science of agents
- ScienceWorld
- ScientistOne
- scratchpad
- SDF
- SDK
- Seed-Prover
- seesaw constraint
- segmentation model
- seL4
- selection bias
- Self-Consistency
- Self-CTRL
- self-efficacy
- self-evolution
- self-inoculation
- self-play
- self-preference
- self-preference bias
- self-preservation
- Self-Refine
- self-trained verification
- semantic attack
- semantic collapse
- semantic entropy
- semantic gap
- semantic hijacking
- Semantic Scholar
- Semgrep
- semi-autoregressive
- semi-bluff
- sentence resampling
- SEO
- separation of powers
- sequential analysis
- sequential decomposition attack
- Sequential Monte Carlo
- serializability
- serialization
- SFT
- SGD
- SGLang
- SGO
- SHADE-Arena
- shadow IT
- shadow pass
- shadow workspace
- shallow alignment
- Shannon's chain rule
- Sheeran example
- shell access
- shell escaping
- Shepherd
- SHERLOC
- shock-capturing scheme
- shortcut penalty
- Show-o2
- showdown
- shutdown resistance
- SIA
- side information
- sigma
- sigmoid
- signal detection theory
- SIGSTOP
- sim-to-real
- SIMD
- SimpleRL-Zoo
- Simplicity Bias
- simulatability
- Simulator Fallacy
- single-cell RNA-seq
- single-rater
- singular value decomposition
- sink neurons
- sink token
- sitemap
- situational awareness
- situational rewards
- SKA
- skill library
- SkillAxe
- SkillOpt
- SkillsBench
- sleep-time compute
- sleeper agent
- sliding window attention
- slow update
- Slumbot
- slyp
- Smallville
- SNR-aware filtering
- SOC
- social engineering
- Socratic-SWE
- softmax
- Sokoban
- sorry
- SORRY-Bench
- Sourcegraph
- sparse attention
- sparse autoencoder
- sparse policy selection
- spatial light modulator
- Spearman correlation
- specializable-generalist
- specification gaming
- spectral element method
- spectral graph theory
- speculative decoding
- Spielman
- spot instances
- SpreadsheetBench
- SQL injection
- SQLite
- square-law detection
- SRE
- SRFT
- SSH
- SSIM
- SSMax
- SST
- Stable Diffusion
- stack trace
- stack-to-pot ratio
- Stackelberg game
- Stag Hunt
- stagnation point
- STALE benchmark
- STAR algorithm
- state resolution
- State Stream Transformer
- State-Space Decoherence
- state-space model
- stateful monitoring
- static analyzer
- stationarity
- stealth browser
- steelman
- steering vector
- steganography
- step amplification factor
- step-level scaling law
- STOC
- stochastic environment
- stop token
- stop-gradient
- strategy prior
- StrategyQA
- stream clustering
- structured handoff
- SU-01
- subgoal-driven framework
- submodularity
- sudo
- Sudoku-Extreme
- sufficient statistic
- summarization smoothing
- sunk cost
- superposition
- Supersede
- supersession
- surrogate cost model
- Sutskever
- Sutton
- SVG
- SWE-bench
- SWE-bench Pro
- SWE-Marathon
- SWIFT
- SWITCH
- sycophancy
- Symbiotica
- symbolic execution
- symlink
- SymPy
- synthetic data
- synthetic document fine-tuning
- system call
- system prompt
T
- tag block
- Take-the-Best
- Tallying
- Taobao
- tape exfoliation
- tape rug pull
- task demand
- Task Formatter
- task-appropriateness reward
- tau-bench
- Tau2-Bench
- tau2-bench
- taxonomy of failures
- TBLite
- teacher forcing
- teacher-forcing replay
- telescoping sum
- temperature
- template collapse
- temporal contrast
- temporal grounding
- temporal knowledge drift
- temporal-difference error
- tensor core
- TensorRT-LLM
- Terminal Wrench
- Terminal-Bench
- TerminalBench
- termination poisoning
- test-time scaling
- test-time training
- tetrahedron
- Texas Hold'em
- text-embedding-three-large
- TextCraft
- TextGrad
- TextLDM
- TGI
- thanatosis
- TheAgentCompany
- theory of mind
- theory-in-use
- thief test
- throughput
- Thurston
- tied embeddings
- time-of-check time-of-use
- time-to-first-token
- TLA+
- TLS
- TMEM
- tmpfs
- token
- token compatibility
- Token Signature
- token-wise potential
- tokenizer
- Tongyi DeepResearch
- tool call
- tool compatibility graph
- tool gateway
- tool poisoning
- tool rug pull
- ToolCUA
- ToolSmith
- top-k
- topology monitor
- TPU
- trace analyzer
- trace induction
- TraceFix
- TracIn
- training harness
- trajectory
- trajectory regulation
- Trajectory Regulation Layer
- transcoder
- transformer
- TransformerLens
- transistor
- transmission matrix
- Triton
- Trixi.jl
- TRL
- TRM
- trojan
- truncated backpropagation
- Truth Ratio
- TruthfulQA
- Tulu
- tuned lens
- Turing
- two-dimensional material
- two-phase locking
- type signature
- type-check
- typed certificate
- typed-state overlay
U
V
- vacuous proof
- valence
- validation gate
- value bet
- value function
- value-at-risk
- value-of-information
- van der Waals heterostructure
- variance reduction
- variational autoencoder
- vector database
- VeOmni
- Vera
- verbalizable
- verifiable reward
- verified system
- verifier
- verl
- vibe coding
- VibeServe
- vision-language model
- vision-language-action model
- visual web agent
- vLLM
- Voyager
- VRAM
- VS Code
- vtable
W
- WADD
- wall-clock time
- warm start
- warmup deep credit assignment
- Wasserstein distance
- watermarking
- Wayback Machine
- weak-to-strong
- web origin
- web scraping
- WebArena
- WebAssembly
- WebGames
- WebRL
- WebShop
- weights
- WHAM
- WHO Safety Checklist
- whois
- WideSearch
- Wikidata
- WinDbg
- windowed FIFO scheduling
- Windows registry
- WMDP
- world model
- World Values Survey
- write primitive