ground truth
In plain wordsThe reference answers treated as correct when checking or training an AI system — for example, the confirmed diagnoses that a medical-imaging model's guesses are compared against.
The values accepted as correct for a set of examples, against which a model is trained or its outputs are scored: verified diagnoses, measured distances, transcribed speech, human-assigned labelslabel, n.: The correct answer attached to an example in a dataset — 'cat' on a photo of a cat, 'positive' on an upbeat review — that a learning system is trained to predict.. In some settings ground truth is a physical measurement independent of human judgment, such as lidar depth used to evaluate a stereo-vision system; in most AI datasets it is the best available human or procedural judgment, adopted as correct by convention. The gap matters. Ambiguous items, annotator disagreement and systematic labeling errors mean that measured accuracy can reward agreement with flawed references, and for open-ended generation there is often no single ground truth at all, which is why evaluation turns to rubrics, multiple references, human evaluationhuman evaluation, n.: Judging an AI's outputs by asking people to rate or compare them, because for many tasks — writing, conversation, translation — no automatic score captures quality well. or model judgesLLM-as-a-judge, n.: Using one AI language model to grade or compare the answers of another, as a cheaper and faster stand-in for human reviewers..
Our ground truth is the biopsy result, not the radiologist's original read.
Origin
A remote-sensing and mapping term: by 1972 NASA described ground truth as observations made on the earth's surface to calibrate measurements from aircraft and satellites. Statistics and machine learning borrowed it for any trusted reference data. An older literary sense, 'fundamental truth', is recorded from 1833.
Usage note
Critics argue the term overstates certainty, since many labels are contestable judgments; 'reference labels' or 'gold labels' are sometimes preferred. A golden datasetgolden dataset, n.: A small, carefully checked set of example questions with known good answers, used as a trusted yardstick to test an AI system every time it changes. is a curated set of ground-truth examples kept for evaluation.
Related
- labellabel, n.: The correct answer attached to an example in a dataset — 'cat' on a photo of a cat, 'positive' on an upbeat review — that a learning system is trained to predict.
- golden datasetgolden dataset, n.: A small, carefully checked set of example questions with known good answers, used as a trusted yardstick to test an AI system every time it changes.
- label noiselabel noise, n.: Mistakes in the answers attached to training or test examples — a dog photo marked 'cat' — which can mislead what a model learns and distort how good it seems.
- annotationannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it.
- evaluationevaluation, n.: Testing an AI system to find out what it can and cannot do, how well, and how safely — from exam-style quizzes to hands-on trials and expert reviews.
- benchmarkbenchmark, n.: A fixed, shared test — a set of questions or tasks with an agreed way of scoring — used to compare AI systems with one another and to track progress over time.
- predictionprediction, n.: A model's output for a given input: its best guess at an unknown answer. In AI it need not be about the future; labeling a photo 'cat' counts as a prediction.
Mentioned in
labellabel, n.: The correct answer attached to an example in a dataset — 'cat' on a photo of a cat, 'positive' on an upbeat review — that a learning system is trained to predict., annotationannotation, n.: Adding information to raw data by hand or with software — marking objects in photos, tagging names in sentences, transcribing speech or ranking chatbot answers — so machines can learn from it., golden datasetgolden dataset, n.: A small, carefully checked set of example questions with known good answers, used as a trusted yardstick to test an AI system every time it changes., inter-annotator agreementinter-annotator agreement, n.: A measure of how often different people labeling the same data give the same answer. Low agreement warns that the task is unclear or the labels are unreliable., label noiselabel noise, n.: Mistakes in the answers attached to training or test examples — a dog photo marked 'cat' — which can mislead what a model learns and distort how good it seems., weak supervisionweak supervision, n.: Training AI with labels that are cheap but imperfect, such as rough rules of thumb, partial tags or noisy guesses, instead of carefully hand-checked answers.