beam search
text generation
Every time a language model finishes a forward pass it hands you not a word but a *probability distribution* over all possible next tokens, and a decoding strategy is the rule…
29,087 technical terms and definitions
text generation
Every time a language model finishes a forward pass it hands you not a word but a *probability distribution* over all possible next tokens, and a decoding strategy is the rule…
decoding strategy, greedy decoding, text generation decoding, sequence search
When a language model finishes a forward pass it does not hand you a word. It hands you a probability distribution over its entire vocabulary, and *decoding* is the policy you use…
nucleus sampling, temperature control, top-k sampling, generation quality
When a language model finishes a forward pass it does not hand you a word. It hands you a probability distribution over its entire vocabulary, and *decoding* is the policy you use…
Beam Search Decoding Strategy Introduction & Motivation Beam Search Decoding Strategy addresses a central problem in generative generation and search-quality-optimality synthesis…
beam steering accuracy prediction, beam steering accuracy, steering accuracy prediction, beam steering, steering accuracy, accuracy prediction
Beam Steering Accuracy Prediction with Machine Learning Introduction & Motivation Beam Steering Accuracy Prediction with Machine Learning addresses a central problem in beam…
audio & speech
Beamforming is spatial filtering that combines multi-microphone signals to emphasize target directions - It boosts desired speech while suppressing interference and ambient noise.…
beamforming ic calibration drift prediction, beamforming ic calibration, ic calibration drift, calibration drift prediction, beamforming ic, ic calibration, calibration drift
Beamforming IC Calibration Drift Prediction with Machine Learning Introduction & Motivation Beamforming IC Calibration Drift Prediction with Machine Learning addresses a central…
bear, reinforcement learning advanced
BEAR is an offline RL algorithm that regularizes policy updates to stay close to dataset action distribution - Distribution constraints, often via divergence bounds, control…
failure analysis advanced
Bed-of-nails is a fixture-based board test method using many spring probes that contact dedicated test points - Parallel contact enables rapid continuity and parametric checks…
quality & reliability
Before-After Comparison is a structured measurement approach that quantifies change impact relative to baseline performance - It is a core method in modern semiconductor…
testing
Behavioral Analysis of ML models is the study of model behavior across different input regions, subgroups, and conditions — going beyond aggregate metrics to understand how…
bc, imitation learning
Behavioral Cloning (BC) is the simplest form of imitation learning — treating the expert's demonstrations as a supervised learning dataset and training a policy to predict…
Behavioral Cloning from Expert Demonstrations Introduction & Motivation Behavioral Cloning from Expert Demonstrations addresses a central problem in reinforcement policy learning…
behavioral model fidelity prediction, behavioral model fidelity, model fidelity prediction, behavioral model, model fidelity, fidelity prediction
Behavioral Model Fidelity Prediction with Machine Learning Introduction & Motivation Behavioral Model Fidelity Prediction with Machine Learning addresses a central problem…
Behavioral Neuroscience & Psychology: Neural Basis of Behavior and Mental Function 1. Neural Circuits and Behavioral Control Behavioral neuroscience examines how nervous system…
explainable ai
Behavioral Testing of ML models is a systematic approach to testing model behavior using input-output test cases — inspired by software engineering testing practices, organizing…
beit, bert pre-training of image transformers, computer vision
BEiT (BERT Pre-Training of Image Transformers) is a self-supervised pre-training method for Vision Transformers that adapts BERT's masked language modeling objective to images…
computer vision
BEiT pre-training is the masked image modeling framework that predicts discrete visual tokens from masked patches, analogous to masked language modeling in NLP - by reconstructing…
belief network bayesian inference prediction, belief network bayesian, network bayesian inference, bayesian inference prediction, belief network, network bayesian
Belief Network Bayesian Inference Prediction with Machine Learning Introduction & Motivation Belief Network Bayesian Inference Prediction with Machine Learning addresses a central…
bell inequality test loopholes, bell inequality test, inequality test loopholes, bell inequality, inequality test, test loopholes
Bell Inequality Test Loopholes Introduction & Motivation Bell Inequality Test Loopholes addresses a central problem in quantum process control…
bell inequality violation tests, bell inequality violation, inequality violation tests, bell inequality, inequality violation, violation tests
Bell Inequality Violation Tests Introduction & Motivation Bell Inequality Violation Tests addresses a central problem in quantum-foundation process control…
bell telephone laboratories, at&t bell labs, murray hill, nokia bell labs, western electric research, 1956 consent decree
Bell Labs, formally Bell Telephone Laboratories, was the research and development organization jointly owned by AT&T and its manufacturing subsidiary Western Electric…
1951 transistor symposium, April 1952 symposium, Transistor Technology books, knowledge transfer, Murray Hill, licensee training
The Bell Labs transistor symposium was a pair of intensive technical meetings, a classified one in September 1951 for military and government representatives and an open one…
bellman-ford negative weight prediction, bellman-ford negative weight, negative weight prediction, bellman-ford negative, negative weight, weight prediction
Bellman-Ford Negative Weight Prediction with Machine Learning Introduction & Motivation Bellman-Ford Negative Weight Prediction with Machine Learning addresses a central problem…
bellman-ford negative weights, bellman-ford negative, negative weights
Bellman-Ford Negative Weights Introduction & Motivation Bellman-Ford Negative Weights addresses a central problem in algorithms process control and negative-cycle-detection…
performance, compare
A benchmark is a fixed test set, a fixed scoring rule, and a published leaderboard — a way of forcing every model to answer the exact same questions so that a claimed improvement…
evaluation
Benchmark is a standardized test suite used to compare models under consistent tasks, data, and scoring rules - It is a core method in modern AI evaluation and safety execution…
data contamination, test set leakage, train-test overlap, n-gram overlap, GPT-3 contamination analysis, dynamic benchmarks, LiveBench, memorization
Benchmark contamination is the situation in which a model's training data contains some or all of a benchmark's test questions, answers, or close paraphrases, so that a high score…
evaluation
A benchmark is a fixed test set, a fixed scoring rule, and a published leaderboard — a way of forcing every model to answer the exact same questions so that a claimed improvement…
mmlu, humaneval
A benchmark suite bundles many individual benchmarks together so a model gets scored across a wide range of tasks at once, rather than reporting one narrow number from one single…
design
Benchmarking is the standardized process of measuring and comparing the performance of semiconductor chips, processors, and computing systems using reproducible test workloads…
latency, throughput, ttft, tokens per second, load testing, performance metrics
Benchmarking a large language model specifically has to deal with a problem static multiple-choice benchmarks were never built for: most of what makes one LLM response better than…
benchmark dataset, benchmark datasets, standard evaluation dataset, imagenet, squad, wmt, mmlu, humaneval, test set, leaderboard
A benchmark is a fixed test set, a fixed scoring rule, and a published leaderboard — a way of forcing every model to answer the exact same questions so that a claimed improvement…
quality & reliability
Benefit Realization is the process of verifying that approved improvements produce the expected operational and financial outcomes - It is a core method in modern semiconductor…
Separating Benign from Malicious Anomalies Introduction & Motivation Separating Benign from Malicious Anomalies addresses a central problem in anomaly detection and (target)…
framework, agnostic
BentoML: Unified Model Serving Overview BentoML is an open-source framework for building reliable machine learning serving endpoints. It solves the "It works on my notebook"…
back end of line, back-end-of-line, metal layers
BEOL (Back End of Line) is the interconnect stack built above the transistors that wires everything together — consisting of multiple metal layers (copper, cobalt, tungsten)…
copper interconnect electromigration, copper interconnect reliability, electromigration failure mechanism, beol reliability testing, current density limit interconnect
Electromigration is the diffusion-controlled transport of metal atoms driven by momentum exchange with conducting electrons. At the current densities carried by on-chip…
interconnect resistance, RC delay, metal pitch scaling
BEOL Interconnect Scaling Challenges address the fundamental physics and engineering barriers encountered as metal wire pitch shrinks below 30nm — including exponentially rising…
copper dual damascene, interconnect rc delay optimization, barrier seed deposition, low-k dielectric integration
Copper dual damascene interconnect architectures, electrochemical superfilling, and barrier-seed metallization constitute the back-end-of-line (BEOL) wiring systems that route…
back end of line, interconnect process
BEOL (Back End of Line) — the portion of chip fabrication that creates the multilayer metal interconnect stack connecting transistors to each other and to I/O pads, after…
copper interconnect scaling, beol resistance challenge, dual damascene, narrow pitch metal
Copper dual damascene interconnect architectures, electrochemical superfilling, and barrier-seed metallization constitute the back-end-of-line (BEOL) wiring systems that route…
beol, process integration
BEOL stack is the multilayer interconnect structure from first metal through upper routing and passivation layers - Successive dielectric and metal modules build global wiring…
Bernoulli trial, binary outcome, Jacob Bernoulli, Ars Conjectandi 1713, success probability, coin flip
The Bernoulli distribution is the probability distribution of a single trial with exactly two possible outcomes, conventionally coded 1 for success and 0 for failure, where…
berry curvature gauge field momentum space, quantum anomalous hall effect 2d channel, topological insulator band structure gaafet, zak phase su 2 gauge potential physics
Berry Phase Mechanics, Quantum Berry Curvature Tensors, and Topological Transport Kinetics in Next-Generation 2D Transistor Channels Executive Summary As semiconductor device…
Bidirectional Encoder Representations from Transformers, Devlin 2018, masked language modeling, MLM, encoder-only transformer, fine-tuning, WordPiece, RoBERTa, DistilBERT
BERT (Bidirectional Encoder Representations from Transformers) is a language model released by Jacob Devlin, Ming-Wei Chang, Kenton Lee and Kristina Toutanova at Google in late…
masked language model mlm, bert pretraining, next sentence prediction, bert fine tuning
BERT (Bidirectional Encoder Representations from Transformers) is the influential self-supervised pretraining approach that learns bidirectional contextual representations via…
bert, bidirectional encoder representations, foundation model
BERT (Bidirectional Encoder Representations from Transformers) is a foundational language model introduced by Google in 2018 that revolutionized natural language processing…
bert bidirectional encoder representations from transformers, bert bidirectional encoder representations transformers, bert bidirectional, encoder representations transformers
BERT - Bidirectional Encoder Representations from Transformers Introduction & Motivation BERT: bidirectional pre-training via masked language modeling. Contextual word…
bert bidirectional encoding representation and masked language modeling, bert bidirectional encoding, bidirectional encoding representation, encoding representation masked
BERT Bidirectional Encoding Representation and Masked Language Modeling Introduction & Motivation BERT Bidirectional Encoding Representation and Masked Language Modeling addresses…