Direct Preference Optimization
dpo, rlhf
Direct Preference Optimization (DPO) is a method that aligns large language models to human preferences without requiring a separate reward model — simplifying the RLHF pipeline…
1,686 technical terms and definitions
dpo, rlhf
Direct Preference Optimization (DPO) is a method that aligns large language models to human preferences without requiring a separate reward model — simplifying the RLHF pipeline…
dpo, preference learning, reward free alignment
Direct Preference Optimization (DPO) is a stable, reward-model-free alternative to RLHF that directly optimizes LLM policy on preference data — achieving comparable alignment…
rlhf alternative, preference learning llm, offline preference optimization, dpo loss function
Direct Preference Optimization (DPO) is the simplified alignment technique that trains language models to follow human preferences without requiring a separate reward model…
dpo training, dpo vs rlhf, offline preference learning, reference model dpo
Direct Preference Optimization (DPO) is the alignment training algorithm that optimizes language models directly on human preference data without requiring a separate reward model…
rlhf alternative, preference alignment, reward model free, offline preference learning
Direct Preference Optimization (DPO) is the alignment technique that trains language models to follow human preferences directly from preference pair data without requiring…
direct preference, optimization llm alignment, 1d ai dpo gpt hh higher, direct preference optimization & llm alignment, direct preference optimization, preference optimization llm, llm alignment
Direct Preference Optimization & LLM Alignment Introduction & Motivation Aligning large language models with human preferences, so that they produce outputs judged helpful…
device physics
Direct Tunneling is the quantum mechanical transmission of carriers through the full width of a thin insulating barrier — occurring at low voltages in ultra-thin gate dielectrics…
advanced packaging
Direct Wafer Bonding (often referred to as Fusion Bonding) is the pinnacle of modern semiconductor substrate engineering, representing the miraculous physical and chemical process…
Directed Graph Neural Network Propagation Introduction & Motivation Directed Graph Neural Network Propagation addresses a central problem in graph network representation…
time series models
Directed Information is information-theoretic measure of time-directed dependence and causal information flow. - It distinguishes directional influence from symmetric association…
directed self-assembly, DSA lithography, block copolymer patterning, graphoepitaxy, chemoepitaxy
Directed self-assembly. uses the thermodynamic microphase separation of block copolymers to form dense nanoscale domains whose orientation and placement are constrained…
directed self-assembly defect density prediction, directed self-assembly defect, self-assembly defect density, defect density prediction, directed self-assembly
Directed Self-Assembly Defect Density Prediction with Machine Learning Introduction & Motivation Directed Self-Assembly Defect Density Prediction with Machine Learning addresses…
directed self-assembly defectivity prediction, directed self-assembly defectivity, self-assembly defectivity prediction, directed self-assembly, defectivity prediction
Directed Self-Assembly Defectivity Prediction with Machine Learning Introduction & Motivation Directed Self-Assembly Defectivity Prediction with Machine Learning addresses…
block copolymer lithography, dsa patterning, self assembly semiconductor, dsa defectivity
Directed Self-Assembly (DSA) is the next-generation patterning technique that uses the thermodynamic self-organization of block copolymer molecules to create sub-10 nm features…
block copolymer lithography, dsa grapho-epitaxy, bcp lamellar cylinder phase, dsa pattern placement error
Directed Self Assembly DSA Patterning is a materials-driven lithography technique exploiting block copolymer phase-separation physics to self-organize nanostructures at resolution…
block copolymer lithography, dsa patterning, self aligned patterning, bcp lithography, chemoepitaxy graphoepitaxy, flory huggins parameter
Directed Self-Assembly (DSA) of block copolymers (BCP) is the lithographic patterning technology that exploits thermodynamic microphase separation of diblock copolymer thin films…
block copolymer phase separation, bcp cylinder lamellae, graphoepitaxy chemoepitaxy, dsa contact hole shrink
Directed Self-Assembly (DSA) is block copolymer micro-phase separation guided by chemical/topographic patterns to create sub-lithographic resolution features for contact hole…
directed self assembly pattern defect prediction with machine learning, directed self assembly pattern defect prediction machine learning, directed self, dsa
Directed Self-Assembly Pattern Defect Prediction with Machine Learning Introduction & Motivation Directed Self-Assembly Pattern Defect Prediction with Machine Learning addresses…
block copolymer lithography, dsa pattern rectification, chemoepitaxy graphoepitaxy, sub-lithographic feature formation
Directed Self-Assembly DSA Patterning — Directed self-assembly leverages the thermodynamic self-organization of block copolymer materials to create sub-lithographic features…
directed test generation, directed test, test generation
Directed Test Generation Introduction & Motivation Directed Test Generation addresses a central problem in verification process control and generation-efficiency-cycles…
DirectLiNGAM Non-Gaussian Linear Causal Discovery Introduction & Motivation DirectLiNGAM Non-Gaussian Linear Causal Discovery addresses a central problem in causal inference…
directx 12, d3d12, direct3d, dxr, directml
DirectX definition and practical boundary. is Microsoft’s multimedia API family, with Direct3D 12 providing explicit low-level graphics and compute control for Windows and Xbox…
Dirichlet Process Mixture Models for Non-Parametric Clustering Introduction & Motivation Dirichlet Process Mixture Models for Non-Parametric Clustering addresses a central problem…
dirichlet processes nonparametric bayesian methods, dirichlet processes, nonparametric bayesian methods, crp dp hdp mcmc, dirichlet processes nonparametric, processes nonparametric
Dirichlet Processes and Nonparametric Bayesian Methods Introduction & Motivation Dirichlet processes enable flexible, non-parametric Bayesian modeling without specifying cluster…
time series models
DirRec Strategy is hybrid direct-recursive forecasting combining horizon-specific models with chained predicted features. - It balances direct horizon specialization…
composable infrastructure, memory pooling fabric, resource disaggregation cxl, rack scale disaggregation
Disaggregated Computing Architecture — System designs that decouple compute, memory, storage, and networking resources into independent pools connected by high-speed fabrics…
disaggregated system architecture and network fabric, disaggregated system architecture, system architecture network, architecture network fabric, disaggregated system
Disaggregated System Architecture and Network Fabric Introduction & Motivation Disaggregated System Architecture and Network Fabric addresses a central problem in architecture…
design
Disaggregation is the semiconductor design strategy of decomposing a monolithic system-on-chip (SoC) into multiple smaller, independently designed and manufactured chiplets…
reinforcement learning advanced
Disagreement exploration is an exploration strategy that rewards state-action regions where model or predictor ensemble disagreement is high - Prediction disagreement acts as…
dr plan, recovery time objective, rto, recovery point objective, rpo, warm standby, pilot light
Disaster recovery restores critical services and data after catastrophic site loss, widespread corruption, destructive attack or prolonged control-plane failure. High availability…
disaster recovery cloud, disaster recovery, recovery cloud
Disaster Recovery Cloud Introduction & Motivation Disaster Recovery Cloud addresses a central problem in cloud process control and rto-target-minutes prediction: how to Predict…
disaster recovery planning, disaster recovery, recovery planning
Disaster Recovery Planning Introduction & Motivation Disaster Recovery Planning addresses a central problem in ml-ops process control and rto-target-hours prediction: how…
nlp
Discourse analysis uses NLP to analyze text structure, coherence, and organization — examining how sentences connect, how topics develop, and how texts achieve communicative…
discourse analysis coherence, discourse analysis, analysis coherence
Discourse Analysis Coherence Introduction & Motivation Discourse Analysis Coherence addresses a central problem in nlp process control and discourse-structure-rhetorical-relation…
nlp
Discourse Marker Prediction is a self-supervised pre-training task where the model must predict missing or masked discourse markers (e.g., "however", "therefore", "because")…
nlp
Discourse relation recognition uses NLP to identify how sentences relate to each other — detecting relationships like cause-effect, contrast, elaboration, and temporal sequence…
generative models
Discrete Diffusion models are generative models that apply the diffusion framework to discrete data (tokens, categories, graphs) — instead of adding Gaussian noise to continuous…
discrete diode transistor package lead defect prediction machine learning, discrete diode, transistor package lead, avi, discrete diode transistor
Discrete Diode and Transistor Package Lead Defect Prediction with Machine Learning Introduction & Motivation Discrete Diode and Transistor Package Lead Defect Prediction…
digital manufacturing
Discrete Event Simulation (DES) in semiconductor manufacturing is a simulation methodology that models fab operations as a sequence of events occurring at discrete time points…
discrete logarithm problem, discrete logarithm, logarithm problem
Discrete Logarithm Problem Introduction & Motivation Discrete Logarithm Problem addresses a central problem in cryptography process control and hardness-computational-assumption…
multimodal ai
Discrete Representation is encoding data into finite symbolic or codebook-based units instead of continuous vectors - It simplifies compression, reasoning, and cross-modal…
discrete wavelet transform (dwt) prediction, discrete wavelet transform, wavelet transform (dwt), transform (dwt) prediction, discrete wavelet, wavelet transform, transform (dwt)
Discrete Wavelet Transform (DWT) Prediction with Machine Learning Introduction & Motivation Discrete Wavelet Transform (DWT) Prediction with Machine Learning addresses a central…
manufacturing operations
Discriminant Analysis is a supervised classification method that separates predefined process states using optimal decision boundaries - It is a core method in modern…
metrology
Discrimination (or resolution) in metrology is the smallest change in a measured value that the measurement system can detect — the minimum increment that the gage can…
quality & reliability
Discrimination Ratio is a metric that quantifies how well a measurement system distinguishes part-to-part variation from measurement noise - It indicates whether collected data is…
Discriminative Active Learning DAL via Domain Adversary Introduction & Motivation Discriminative Active Learning DAL via Domain Adversary addresses a central problem in {subject}…
fine-tuning
Discriminative Fine-Tuning is the combined strategy of using different learning rates for different layers (layer-wise LR) during fine-tuning — a term coined by the ULMFiT paper…
healthcare ai
Disease Prediction from Text is the clinical NLP task of inferring likely diagnoses or disease risk from unstructured clinical narratives, patient-reported symptoms, and medical…
healthcare ai
Disease progression modeling uses machine learning to predict how diseases evolve over time — analyzing longitudinal patient data to forecast symptom trajectories, functional…
Disentangled Attention is the core attention mechanism of DeBERTa that separates token content and position into independent vectors — computing three types of attention…