Alphabetical Index of Terms
One line per concept, with every record of it in the collection. "See" points from a variant name, abbreviation or merged occurrence to the concept's headword; "compare" and other stated relations mark terms that are related but not synonyms. † marks a provisional concept identity; ◆ a concept new for the 2025 Collection. In the PDF edition the numbers after each concept are page references; online, each record ID links to its entry.
"See also" after a concept lists the concepts its entries name in their own See Also sections, as written (one direction only; harvested by Muse Spark 1.3). Aliases checked by Muse Spark 1.3, November 14, 2026: pass.
- Advanced Reasoning Models (Cognitive Hybrid Architectures) † —
AIK-2025-S-P-ADVREASON - Agentic AI —
AIK-2025-E-P-AGENTICAI,AIK-2025-E-P-AGENTICAI-02 - Agentic AI / Super-Agents / Model Orchestration → see Agentic AI
- AgentScen — LLM-Based External World Models —
AIK-2025-S-P-AGENTSCEN - AI Alignment —
AIK-2025-S-P-ALIGNMENT - AI Hallucination —
AIK-2025-S-P-HALLUC - AI Washing —
AIK-2025-S-P-AIWASHING - Alignment Faking (Deceptive Alignment) —
AIK-2025-E-P-ALIGNFAKE - Alignment Patterns: Constitutional, Preference-Modeling, & Red-Teaming —
AIK-2025-S-P-ALIGNPATTERN - Arithmetic and Reasoning Heads † —
AIK-2025-S-P-ARITHHEADS - Artificial Hivemind —
AIK-2025-E-P-HIVEMIND - Bidirectional Human-AI Alignment —
AIK-2025-S-P-BIDIRALIGN - CALM → see Continuous Autoregressive Language Models (CALM)
- Chain-of-Thought Monitorability ◆ —
AIK-2025-F-N-COTMONITOR· See also: Chain-of-Thought Prompting; Context Engineering; Context Rot; Emergent Misalignment; Scheming (AI) - Chain-of-Thought Prompting —
AIK-2025-F-P-COTPROMPT· See also: Chain-of-Thought Monitorability; Chain-of-Verification (CoVe); Test-Time Compute Scaling - Chain-of-Verification (CoVe) —
AIK-2025-S-P-COVE - Circuit Sparsity —
AIK-2025-F-P-CIRCUITSP· See also: Mechanistic Interpretability Tools; Mixture of Experts (MoE) - Composable Models & Model Chaining → see Compound AI Systems
- Compound AI Systems —
AIK-2025-E-P-COMPOUNDAI,AIK-2025-S-P-COMPOSABLE - Confident Adaptive Language Modeling (Schuster et al., 2022) — a different method with the same acronym (Editorial Addition E01); compare Continuous Autoregressive Language Models (CALM)
- Constitutional AI —
AIK-2025-F-P-CONSTAI,AIK-2025-S-P-CONSTAI· See also: AI Alignment; RLAIF — Reinforcement Learning from AI Feedback - Constitutional AI 2.0 † —
AIK-2025-S-P-CAI2 - Context Engineering ◆ —
AIK-2025-F-N-CONTEXTENG· See also: Agentic AI; Chain-of-Thought Prompting; Model Context Protocol (MCP); RLVR (Reinforcement Learning with Verifiable Rewards) - Context Rot ◆ —
AIK-2025-F-N-CONTEXTROT· See also: Compound AI Systems; Context Engineering; Long-Context & Windowing Techniques; Retrieval-Augmented Generation (RAG); Test-Time Compute Scaling - Continual World-Model Agents → see EvoAgent — Continual World-Model Agents
- Continuous Autoregressive Language Models (CALM) —
AIK-2025-F-P-CALM· See also: Mixture of Experts (MoE); Speculative Decoding - Controllable Generation Frameworks —
AIK-2025-S-P-CTRLGEN - CoT monitorability → see Chain-of-Thought Monitorability
- CoT prompting → see Chain-of-Thought Prompting
- CoV (historical abbreviation; corrected to CoVe) → see Chain-of-Verification (CoVe)
- CoVe → see Chain-of-Verification (CoVe)
- Deceptive Alignment — broader concept, discussed under Alignment Faking (Deceptive Alignment)
- Delta Learning Hypothesis —
AIK-2025-S-P-DELTALEARN - Dexterous Manipulation → see Soft Robotics / Dexterous Manipulation
- Distillation at Scale / Student-Teacher Cascades —
AIK-2025-S-P-DISTILSCALE - Drop-Upcycling (Sparse MoE Training) —
AIK-2025-S-P-DROPUPCYCLE - Dynamic Planning for LLM Agents —
AIK-2025-S-P-DYNPLAN - Dynamical Regularization (in Diffusion) —
AIK-2025-S-P-DYNREG - E-Processes / Game-Theoretic Statistics —
AIK-2025-E-P-EPROCESS - Edge AI —
AIK-2025-S-P-EDGEAI,AIK-2025-S-P-EDGEAI-02 - Embodied AI (as a Paradigm) —
AIK-2025-S-P-EMBODIEDAI - Embodied Foundation Models —
AIK-2025-S-P-EMBODIEDFM - Emergent Misalignment ◆ —
AIK-2025-F-N-EMERGENTMIS· See also: Chain-of-Thought Monitorability; Constitutional AI; Context Engineering; Context Rot; RLVR (Reinforcement Learning with Verifiable Rewards); Scheming (AI) - EvoAgent — Continual World-Model Agents —
AIK-2025-S-P-EVOAGENT - Explainable AI (XAI) —
AIK-2025-S-P-XAI - Federated & On-Device Model Training —
AIK-2025-S-P-FEDONDEVICE - Frugal AI —
AIK-2025-S-P-FRUGALAI,AIK-2025-S-P-FRUGALAI-02 - Game-Theoretic Statistics → see E-Processes / Game-Theoretic Statistics
- Gated Attention —
AIK-2025-E-P-GATEDATTN - Generative Engine Optimisation (GEO) —
AIK-2025-S-P-GEO - Generative Feedback Loop (Self-Adaptation) —
AIK-2025-S-P-GENFEEDBACK - Generative User Interfaces (GenUI) —
AIK-2025-F-P-GENUI - GenUI → see Generative User Interfaces (GenUI)
- GEO → see Generative Engine Optimisation (GEO)
- GKD → see On-Policy Distillation
- GraphRAG —
AIK-2025-S-P-GRAPHRAG - High-Dimensional Learning Dynamics —
AIK-2025-S-P-HDLD - Hyperautomation —
AIK-2025-S-P-HYPERAUTO - KV-Cache Sharing (On-Device Inference) —
AIK-2025-E-P-KVSHARE - Large Reasoning Models (LRMs) —
AIK-2025-S-P-LRM,AIK-2025-E-P-LRM - Large-Scale Population Agent Simulation (LLM Archetypes) —
AIK-2025-S-P-POPAGENTSIM - Layer-Adaptive Expert Inference → see LExI — Layer-Adaptive Expert Inference
- LExI — Layer-Adaptive Expert Inference —
AIK-2025-S-P-LEXI - LLM-Based External World Models → see AgentScen — LLM-Based External World Models (the record of a 2025 workshop; for world models as a method, see also World Model (LLM-Integrated))
- Long-Context & Windowing Techniques —
AIK-2025-S-P-LONGCONTEXT - Long-Horizon LLM Planning & Efficiency Metrics † —
AIK-2025-S-P-LHPLANMETRIC - LRMs → see Large Reasoning Models (LRMs)
- MCP → see Model Context Protocol (MCP)
- Mechanistic Interpretability Tools —
AIK-2025-E-P-MECHINTERP - Memorization Sinks —
AIK-2025-S-P-MEMSINK - Mixture of Depths —
AIK-2025-F-P-MOD· See also: Mixture of Experts (MoE) - Mixture of Experts (MoE) —
AIK-2025-F-P-MOE,AIK-2025-S-P-MOE,AIK-2025-S-P-MOE-02 - Mixture of Latent Experts (MoLE / MoLAE) —
AIK-2025-S-P-MOLE - Mixture-of-Experts (MoE) — sparse expert routing → see Mixture of Experts (MoE)
- Model Context Protocol (MCP) —
AIK-2025-S-P-MCP,AIK-2025-E-P-MCP - Model Orchestration — named in the Launch Edition headword with Agentic AI (not a synonym); discussed under Agentic AI
- Model Safety Evaluation Suites & LLM-Based Evaluators —
AIK-2025-S-P-SAFETYEVAL - MoE → see Mixture of Experts (MoE)
- MoLAE → see Mixture of Latent Experts (MoLE / MoLAE)
- MoLE → see Mixture of Latent Experts (MoLE / MoLAE)
- Multi-Agent Debate Systems —
AIK-2025-S-P-MADEBATE - Multimodal Native Architecture —
AIK-2025-S-P-MMNATIVE - Nested Learning (Continual AI) —
AIK-2025-S-P-NESTEDLEARN - On-Policy Distillation —
AIK-2025-F-P-ONPOLDISTIL· See also: Chain-of-Thought Prompting; RLVR (Reinforcement Learning with Verifiable Rewards) - Parameter-Efficient Fine-Tuning (PEFT) evolution —
AIK-2025-S-P-PEFT - PEFT → see Parameter-Efficient Fine-Tuning (PEFT) evolution
- Plan-and-Act Agent Framework —
AIK-2025-S-P-PLANACT - Prefill–Decode Disaggregation —
AIK-2025-F-P-PDDISAGG· See also: KV-Cache Sharing (On-Device Inference); Mixture of Experts (MoE); Speculative Decoding - Programmatic Representations (for Agents) —
AIK-2025-S-P-PROGREP - Prompt Injection Defense Layers —
AIK-2025-S-P-PIDEFENSE - Quantization & Ultra-Low-Precision Inference —
AIK-2025-S-P-QUANTINFER - RAG → see Retrieval-Augmented Generation (RAG)
- Reinforcement Learning from AI Feedback → see RLAIF — Reinforcement Learning from AI Feedback
- Reinforcement Learning with Verifiable Rewards → see RLVR (Reinforcement Learning with Verifiable Rewards)
- Retrieval-Augmented Generation (RAG) —
AIK-2025-S-P-RAG,AIK-2025-S-P-RAG-02 - Retrieval-Augmented Generation (RAG) — next-gen → see Retrieval-Augmented Generation (RAG)
- RLAIF — Reinforcement Learning from AI Feedback —
AIK-2025-S-P-RLAIF,AIK-2025-S-P-RLAIF-02 - RLVR (Reinforcement Learning with Verifiable Rewards) ◆ —
AIK-2025-F-N-RLVR· See also: Large Reasoning Models (LRMs); On-Policy Distillation; RLAIF — Reinforcement Learning from AI Feedback; Test-Time Compute Scaling - Scalable Oversight & Synthetic Critiques —
AIK-2025-E-P-SCALEOVER,AIK-2025-S-P-SCALEOVER - Scalable Oversight via Synthetic Critiques → see Scalable Oversight & Synthetic Critiques
- Scaling Laws for Sparse Models & Efficiency Metrics —
AIK-2025-S-P-SPARSESCALE - Scheming (AI) ◆ —
AIK-2025-F-N-SCHEMING· See also: Alignment Faking (Deceptive Alignment); Chain-of-Thought Monitorability; Constitutional AI; Emergent Misalignment - Self-Supervised RL —
AIK-2025-S-P-SSRL - Shadow AI —
AIK-2025-S-P-SHADOWAI - Shared Autonomy (Robotics) —
AIK-2025-S-P-SHAREDAUTO - SLMs → see Small Language Models (SLMs)
- Small Language Models (SLMs) —
AIK-2025-S-P-SLM - Social Choice Fairness —
AIK-2025-S-P-SOCCHOICE - Soft Robotics / Dexterous Manipulation —
AIK-2025-S-P-SOFTROBOT - Speculative Decoding —
AIK-2025-S-P-SPECDECODE - Student-Teacher Cascades → see Distillation at Scale / Student-Teacher Cascades
- Super-Agents — named in the Launch Edition headword with Agentic AI (not a synonym); discussed under Agentic AI
- Synthetic Data —
AIK-2025-S-P-SYNDATA - Synthetic Data Quality Filtering —
AIK-2025-S-P-SYNDATAFILT - Test-Time Compute Scaling —
AIK-2025-E-P-TTCSCALE - Theoretical Learning Dynamics of MoE † —
AIK-2025-S-P-MOETHEORY - Tokenization-Centric Design —
AIK-2025-S-P-TOKDESIGN - Vector Databases —
AIK-2025-S-P-VECTORDB - Vertical AI Integration —
AIK-2025-S-P-VERTICALAI - Vibe Coding —
AIK-2025-S-P-VIBECODE - World Model (LLM-Integrated) —
AIK-2025-E-P-WORLDMODEL,AIK-2025-S-P-WORLDMODEL - World Models → see World Model (LLM-Integrated)
- XAI → see Explainable AI (XAI)