Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
An LLM (large language model) is a neural network that generates text by predicting the next token from the context it receives. Most current LLMs use Transformer architecture. For developers, the important consequences are practical: tokens and context windows limit requests, embeddings make semantic retrieval possible, prompting changes runtime instructions, fine-tuning changes model parameters, and retrieval-augmented generation (RAG) supplies external information at runtime. Reliable applications also need evaluations, access controls, and defenses against prompt injection.
This glossary explains those terms in the order they affect an AI application: how a model represents text, how it receives context, how you adapt it, and how you measure and secure the result.
What is an LLM and how does it generate text?
A large language model is a text-focused foundation model trained on a very large corpus. Its learned weights encode statistical and semantic patterns. During generation, the model predicts a probability distribution for the next token, selects a token according to its decoding settings, adds it to the context, and repeats the process until it reaches a stopping condition or output limit.
Generative AI
Generative AI is the broader category of systems that create new content, including text, images, audio, code, or video. An LLM is the language-and-text member of that family; calling every generative model an LLM is inaccurate.
#1 Best Overall
- Read Before You Buy — No Video Output: These adapters support charging and USB 2.0 data transfer, but cannot transmit video signals. Except for standard USB webcams (which use USB data only), they are not compatible with HDMI/DisplayPort cables, video-capable USB-C hubs, or docking stations with video output.
- Convert USB-A Ports to USB-C: Designed to connect USB-C earphones, cables, flash drives, card readers, and other USB-C accessories to standard USB-A ports. Plug-and-play with no drivers or software required.
- Aluminum Alloy Housing: Built with a sturdy aluminum alloy shell that aids in heat dissipation and protects against daily wear and scratches. Designed to maintain a stable and secure connection.
- Compact & Travel-Friendly: The ultra-compact design allows the adapter to stay plugged into your device without blocking adjacent ports or adding bulk, reducing wear and tear on your original USB ports.
- 12-Month Warranty: Backed by a 12-month manufacturer warranty for peace of mind. Designed to meet strict quality control standards for reliable everyday performance.
Transformer
A Transformer is the dominant neural-network architecture for modern language models. Its self-attention mechanism lets each token weigh information from other tokens in the supplied sequence, helping the model connect words and concepts that are far apart. Implementations differ: do not assume every production model has the same encoder, decoder, or attention layout.
Parameters, weights, and pretraining
A parameter (also called a weight) is a learned numerical value. The collection of weights stores patterns acquired during training. Pretraining is the initial, usually large-scale training phase; an autoregressive model commonly learns by predicting the next token from preceding tokens, even when the source corpus has no human labels.
Inference
Inference means running the trained model on an input to produce a prediction or generated response. In an application, inference is the serving-time operation, separate from pretraining or fine-tuning. Its latency and cost depend on the model, input and output lengths, hardware, batching, and service configuration.
What is a token, and why does a context window matter?
A token is the unit a model processes. Depending on the tokenizer, a token may represent a character, a whole word, part of a word, punctuation, or whitespace. Tokenization converts text into the model’s vocabulary items and token IDs; the vocabulary is the set of IDs that tokenizer can emit.
The same sentence can produce different token counts on different models because each model can use a different vocabulary and tokenizer. Token counts affect request limits, usage accounting, prompt design, and latency, so measure with the tokenizer for the model you actually call rather than estimating from characters or words.
Context window
A context window is the amount of tokenized text a model can look back at and process for one request. It is working memory for that request, not the model’s complete training corpus and not automatically a permanent memory store. The prompt, conversation history, retrieved documents, tool results, and requested output all consume model-specific limits; verify current input and output limits in the documentation for the chosen model.
Design implications
- Count tokens before sending large documents or long conversation histories.
- Reserve space for the answer when setting an output limit.
- Trim, summarize, or retrieve only relevant history when a context window is nearly full.
- Expect longer contexts to affect latency and potentially cost, even when they fit the formal limit.
What are embeddings and vector search used for?
An embedding is a vector representation of text or another input, designed so useful relationships are reflected in geometric distance. Similar meanings tend to produce nearby vectors, although “nearby” is meaningful only for the embedding model and similarity method being used.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Embeddings support semantic search, clustering, recommendations, anomaly detection, and classification. In a retrieval system, an application embeds documents, stores those vectors, embeds a user query, and searches for nearby document vectors.
Rank #2
- 5-in-1 USB-C Hub: Experience comprehensive connectivity featuring a Power Delivery input, two USB-A 2.0 ports, a USB-A 3.0 port, and an HDMI port. (Note: The USB-C power delivery input port is only for connecting an external wall charger to power your laptop and cannot power peripheral devices.)
- 90W Pass-Through Charging: Achieve optimal charging with 90W pass-through power to your laptop, supported by a total input of 100W, with the hub reserving 10W for operational efficiency. (Note: Wall charger not included.)
- Quick Data Transfers: Accelerate your productivity with rapid data transfers using a high-speed 5Gbps USB 3.0 port and two 480Mbps USB 2.0 ports.
- 4K HDMI Display: Enhance your visual experience with a hub capable of delivering 4K resolution at 30Hz in both mirror and extend modes. Please note that this hub is compatible with MacBook (macOS 12 and newer), Windows 10 and 11, ChromeOS, and laptops equipped with DP Alt Mode and Power Delivery. Note: This device is not compatible with Linux.
- What You Get: Anker USB-C Hub (5-in-1, 4K HDMI), welcome guide, 18-month warranty, and our friendly customer service.
Vector database and similarity search
A vector database stores embedding vectors and returns nearby vectors for a query vector. It is an implementation component, not the embedding model itself. A production search may combine vector similarity with keyword filters, metadata filters, permissions, recency, or a reranking step.
Embedding quality, chunk boundaries, metadata, and access filtering determine what a model can retrieve. A sophisticated generator cannot answer from a document that the retriever failed to return.
How do prompting and instruction layers work?
A prompt is the input supplied at inference time: instructions, examples, user data, conversation state, and sometimes retrieved documents or tool results. Prompts are not the model’s training data and do not alter its weights.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →System, developer, and user instructions
Many APIs expose separate system, developer, and user messages. These layers commonly express application-wide behavior, implementation constraints, and the user’s request, respectively. Precedence and conflict handling are defined by the particular API or framework; do not assume all vendors implement identical roles. Enforce critical permissions in application code rather than relying on a message hierarchy.
Prompt engineering
Prompt engineering is the deliberate design and iteration of inputs to obtain a desired result. A robust prompt specifies the task, relevant context, constraints, success criteria, and output format. Examples can show edge cases and formatting more reliably than a vague request.
- Zero-shot prompting: asks for the task without an example.
- One-shot prompting: includes one example.
- Few-shot prompting: includes several examples and is a form of in-context learning.
Prompt tuning is not prompt writing
Prompt tuning is a parameter-efficient adaptation method that learns a soft prompt or prefix while leaving most model parameters frozen. Ordinary prompt engineering edits human-readable text at runtime; prompt tuning trains learned vectors.
Temperature and context engineering
Temperature is a decoding control that generally makes outputs more varied at higher values and more deterministic at lower values. Exact behavior is implementation-specific, and temperature does not make an incorrect answer factual. Inference-time context engineering is the broader practice of selecting and ordering instructions, user data, retrieved passages, conversation state, and tool results so the model receives the right information in the right form.
Recommended Free Tools
What is RAG, and how does grounding reduce hallucinations?
Retrieval-augmented generation (RAG) is a three-stage pattern: retrieve relevant data, augment the model prompt with that data, and generate a response. It can provide current, proprietary, or domain-specific information that was absent from pretraining without changing model weights.
Rank #3
- Sleek 7-in-1 USB-C Hub: Features an HDMI port, two USB-A 3.0 ports, and a USB-C data port, each providing 5Gbps transfer speeds. It also includes a USB-C PD input port for charging up to 100W and dual SD and TF card slots, all in a compact design.
- Flawless 4K@60Hz Video with HDMI: Delivers exceptional clarity and smoothness with its 4K@60Hz HDMI port, making it ideal for high-definition presentations and entertainment. (Note: Only the HDMI port supports video projection; the USB-C port is for data transfer only.)
- Double Up on Efficiency: The two USB-A 3.0 ports and a USB-C port support a fast 5Gbps data rate, significantly boosting your transfer speeds and improving productivity.
- Fast and Reliable 85W Charging: Offers high-capacity, speedy charging for laptops up to 85W, so you spend less time tethered to an outlet and more time being productive.
- What You Get: Anker USB-C Hub (7-in-1), welcome guide, 18-month warranty, and our friendly customer service.
- Index: split authorized source material, create embeddings, and store vectors with metadata.
- Retrieve: embed the query and return relevant passages, applying permission and metadata filters.
- Augment: place the selected passages and their source information in the model’s context.
- Generate: instruct the model to answer from the supplied evidence and to acknowledge when evidence is insufficient.
Grounding connects an answer to trusted or retrieved evidence and can support source attribution. It improves reliability but is not a correctness guarantee: retrieval can miss the right passage, a source can be wrong, and the model can misread or overstate evidence.
Hallucination
A hallucination is an output that sounds plausible but is unsupported or factually wrong. RAG, citations, constrained formats, and verification checks reduce this failure mode; none proves that every generated statement is true.
Tool use, function calling, and agents
Tool use or function calling lets a model request that application code or an external service perform a structured operation. Validate arguments, authenticate the caller, authorize the action, and check the result in code; model-generated text is not a security boundary.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteAn AI agent is the surrounding application that combines a model with orchestration, state or memory, and tools to pursue a goal. The agent is not merely the underlying model. More tools and autonomy increase the need for limits, logging, and human approval for consequential actions.
What is the difference between prompting, RAG, and fine-tuning?
Choose among them by asking where the needed knowledge lives and how often it changes. Prompting changes the request, RAG changes the runtime evidence, and fine-tuning changes the model’s learned parameters.
| Approach | Where knowledge or behavior lives | Update path | Main operational concerns | What to evaluate |
|---|---|---|---|---|
| Prompting | Instructions and examples in the context | Edit the prompt or message templates | Context length, instruction conflicts, and runtime cost | Instruction adherence and task quality |
| RAG | External documents retrieved at runtime | Refresh documents, embeddings, and the index | Retrieval latency and recall, permissions, citations, and prompt injection in retrieved text | Retrieval recall, groundedness, citation quality, and answer quality |
| Fine-tuning | Model parameters or adapter weights | Train an adapted model or adapter on new examples | Training data quality, compute, versioning, and possible data leakage | Task metrics, instruction adherence, safety, and regression against a baseline |
Use RAG when facts are private, changing, or need source attribution. Use fine-tuning when repeated examples show a stable task, style, or behavior that prompts cannot reliably produce. The approaches can be combined: a tuned model can still use RAG, and a carefully designed prompt remains necessary for either.
When should I fine-tune, and what do PEFT and LoRA mean?
Fine-tuning continues training a pretrained model on additional examples so it learns task, domain, or style patterns. It changes model parameters, unlike supplying an instruction in a prompt.
Instruction tuning
Instruction tuning is fine-tuning on instruction-and-response examples to improve the model’s ability to follow natural-language tasks. Quality and coverage of examples matter more than simply increasing their count.
Rank #4
- Dual Converters, Infinite Potential：Includes 2× USB C male to USB A female adapters and 2× USB A male to USB C female adapters. Perfect for a wide range of uses—tablets with Bluetooth keyboards, expand USB ports on macbook, and more. Two different converters for all your daily needs
- Next-Level 10Gbps & 3A Charging: No more slow 480Mbps, this usb to usb c adapter has a transfer speed of up to 10Gbps, allowing you to do more transferring in less time. This usb adapter fits both USB A and USB C charger, supporting up to 3A fast charging
- Upgraded Exquisite Craftsmanship: With an aluminum alloy housing and metal connector, the usbc to usb adapter is extremely durable and sturdy. Rigorously tested to withstand more than 10,000 times of plugging and unplugging, ensuring long-lasting performance
- Broad Compatible: The usb c to usb adapter widely supports all USB C/ USB A devices like laptops, tablets, cellphones, car chargers, and phone chargers. Such as compatible with MacBook Pro/Air 2023/2022, Thunderbolt 4/3 Devices,Apple MagSafe Watch 9/8/7/SE/Ultra, iPad Pro 2022/2021, Samsung Galaxy S23/S20/S10, and iPhone 17/16/15 Pro. Plug and play
- Please Note: To reach 10Gbps speed, keep the cable under 3.3 ft. For USB A Male to USB C adapters, try flipping the USB C connector. USB C Male to USB A adapters support bidirectional 10Gbps transfer within 3.3 ft
Parameter-efficient fine-tuning (PEFT)
PEFT trains only a small subset of parameters or an adapter while keeping most base weights frozen. This reduces compute and storage requirements compared with updating every weight.
LoRA
Low-Rank Adaptation (LoRA) is a PEFT method that trains low-rank update matrices while the base model remains largely frozen. Serving systems can attach multiple adapters to one base model, but each adapter still needs its own evaluation and version controls.
Distillation
Distillation trains a smaller student model to reproduce useful behavior from a larger teacher model. The exact loss function and data recipe vary by implementation, so validate the student on the target workload instead of assuming that teacher-like outputs guarantee equivalent quality.
Quantization and caching
Quantization represents weights or activations with lower numerical precision to reduce memory use and improve serving efficiency, with possible quality trade-offs. Benchmark the selected implementation on representative inputs. Caching reuses previously computed prompts, embeddings, or responses to reduce latency and cost; define cache keys, freshness rules, and privacy boundaries explicitly.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How do I evaluate an AI application?
An LLM evaluation, or eval, is a set of tests, metrics, or benchmarks for task quality, helpfulness, safety, and other requirements. Build it around the real user outcome rather than relying only on a generic benchmark score.
Prompt and response sets
A prompt set is the collection of evaluation inputs. A response set is the model’s output for those inputs. Version both sets so a prompt, model, retrieval, or code change can be compared against a known baseline.
Human evaluation
People can review correctness, relevance, style, or safety using a rubric. Record the rubric, sampling method, evaluator instructions, and disagreements; there is no universal human-rating protocol that fits every application.
Latency and throughput
Latency is the time to produce a response, including time to first token in a streaming system. Throughput is the amount of work completed per unit time. Measure both under representative concurrency, prompt lengths, retrieval behavior, and tool calls rather than quoting a single best-case run.
Best Value
- 5-in-1 Connectivity: Equipped with a 4K HDMI port, a 5 Gbps USB-C data port, two 5 Gbps USB-A ports, and a USB C 100W PD-IN port. Note: The USB C 100W PD-IN port supports only charging and does not support data transfer devices such as headphones or speakers.
- Powerful Pass-Through Charging: Supports up to 85W pass-through charging so you can power up your laptop while you use the hub. Note: Pass-through charging requires a charger (not included). Note: To achieve full power for iPad, we recommend using a 45W wall charger.
- Transfer Files in Seconds: Move files to and from your laptop at speeds of up to 5 Gbps via the USB-C and USB-A data ports. Note: The USB C 5Gbps Data port does not support video output.
- HD Display: Connect to the HDMI port to stream or mirror content to an external monitor in resolutions of up to 4K@30Hz. Note: The USB-C ports do not support video output.
- What You Get: Anker 332 USB-C Hub (5-in-1), welcome guide, our worry-free 18-month warranty, and friendly customer service.
Provenance
Provenance is information about how digital content was created or changed. Retain the document versions, model and adapter versions, prompt template, retrieved sources, tool calls, evaluator results, and relevant timestamps needed to audit an output.
What security terms should every LLM developer know?
Prompt injection
Prompt injection is crafted input that attempts to override intended instructions or manipulate model behavior. Treat user text, retrieved documents, web pages, and tool outputs as untrusted input. Keep secrets out of model-visible context where possible, isolate instructions from data, validate tool arguments, enforce permissions outside the model, and constrain outputs before executing them.
RLHF and preference optimization
Reinforcement learning from human feedback (RLHF) and other preference-optimization methods use human or preference signals to shape model behavior after pretraining. Algorithms, data, and guarantees vary by model, so the acronym alone does not establish a particular safety or quality level.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
HHH
HHH describes three desired properties: helpful assistance, honest behavior that acknowledges limitations, and harmless or non-harmful assistance. It is a behavioral framework, not a proof that a system is safe in every context.
How do these terms fit together in a production system?
- Tokenize and budget the request so instructions, retrieved evidence, history, and output fit the context window.
- Use clear system or developer instructions and representative examples, then choose decoding settings appropriate to the task.
- For changing or private facts, retrieve authorized passages through embeddings and a vector search layer.
- Pass tool requests through typed schemas, authentication, authorization, validation, and logging.
- Run versioned evals for quality, groundedness, safety, latency, and throughput before and after every material change.
- Store provenance and monitor failures such as unsupported claims, retrieval misses, prompt injection, and stale caches.
Frequently Asked Questions
Can a longer context window replace a database or memory system?
No. A context window is temporary working memory for a request, and older conversation content may need to be truncated or summarized. Durable memory requires an application-managed store with explicit retention, retrieval, and deletion rules.
How should I test an LLM feature when there is no single correct wording?
Define acceptance criteria for the outcome, such as required facts, prohibited claims, format, and safety behavior. Combine automated checks with rubric-based human review, and keep difficult edge cases in a versioned regression set.
Should retrieved documents be trusted because they came from my own system?
No. Internal documents can be stale, incorrectly permissioned, or contain instructions aimed at manipulating the model. Apply access controls and treat every retrieved passage as data to be evaluated, not as an instruction with authority.
Free tools Windows power users keep installed
One-click scans. No signup required.
What should I log when a model calls a tool?
Record the model and prompt versions, structured arguments, authorization decision, tool result, timestamps, and the final response. Redact secrets and personal data according to your retention policy, and make logs tamper-evident when actions have regulatory or financial consequences.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
