zero-shot learning
In plain wordsDoing a task without being shown a single example of it — say, an AI sorting reviews into positive and negative just from being told to, or recognizing an animal it never saw in training.
- 1
Language models
Performing a task specified only by a natural-language instruction or description, with no worked examples in the promptprompt, n.: The message or instructions you give an AI system — a question, a request, some examples or a document — that it responds to. Clearer prompts generally get better results. and no task-specific training. Zero-shot ability comes from pretrainingpretraining, n.: The first and usually the most expensive stage of building a modern AI model, in which it learns general patterns from enormous amounts of raw text, images or code before being specialized for particular uses. on diverse text and is greatly strengthened by instruction tuninginstruction tuning, n.: Training a language model on many tasks written as plain instructions with correct answers, so it learns to do what a request asks instead of just continuing the text.; it is the baseline setting against which few-shot promptingfew-shot prompting, n.: Showing an AI a handful of examples of a task — inputs with the answers you want — inside the prompt, so it can copy the pattern for a new case. and fine-tuningfine-tuning, n.: Taking an AI model that has already learned a lot and training it a little more on a smaller, focused set of examples so it gets better at one job or behaves a certain way. are compared. Zero-shot chain of thoughtchain of thought, n.: Getting an AI to write out its reasoning step by step before giving an answer, the way a student shows their working. This often makes answers to math and logic problems more accurate. adds only a generic cue to reason step by step.
Zero-shot, the model labeled support tickets at 88% accuracy; five examples pushed it to 93%.
- 2
Classical machine learning
Classifying examples of categories that were entirely absent from the training data, by way of side information that describes every category, seen or unseen: lists of attributes (aquatic, spotted, four-legged), word embeddingsword embedding, n.: A list of numbers standing for a word, learned so that words used in similar ways get similar lists. It lets a computer measure that 'cat' is closer to 'dog' than to 'car'. of the class names, or short text descriptions. A model that has learned to detect the attributes from images of known species can then name an animal it has never seen, by matching the attributes it detects against each unseen species' description. Contrastive image–text models such as CLIPCLIP, prop. n.: An OpenAI model that learned to match pictures with their captions from hundreds of millions of web examples, so it can judge how well any piece of text describes any image. made zero-shot image classification, by comparing images with class-name prompts, routine.
Origin
Hugo Larochelle, Dumitru Erhan and Yoshua Bengio described 'zero-data learning' in 2008; the name 'zero-shot learning' appears in Mark Palatucci and colleagues' 2009 NIPS paper, alongside Christoph Lampert and colleagues' attribute-based classifiers of the same year. For language models, GPT-2GPT-2, prop. n.: A 2019 OpenAI language model that wrote surprisingly fluent paragraphs. OpenAI at first withheld the full version, saying it could be misused for fake news, which made it famous. (2019) framed tasks as zero-shot, and GPT-3GPT-3, prop. n.: A huge 2020 OpenAI language model that could perform new tasks simply from instructions or a few examples written into its prompt, without retraining. It made prompting a new way of programming. (2020) fixed the zero-, one- and few-shot vocabulary.
Usage note
In the LLM sense, 'zero-shot' says nothing about whether similar tasks appeared in the training data; strong zero-shot results may partly reflect data contaminationdata contamination, n.: When the test questions used to grade an AI system were already in its training data, so a high score may reflect memorized answers rather than real ability..
Related
- few-shot promptingfew-shot prompting, n.: Showing an AI a handful of examples of a task — inputs with the answers you want — inside the prompt, so it can copy the pattern for a new case.
- in-context learningin-context learning, n.: A language model's ability to pick up a new task from instructions or examples placed in its input, on the spot, without any retraining — like a quick study learning a game's rules from one round.
- few-shot learningfew-shot learning, n.: Getting an AI system to handle a new task or recognize a new kind of thing from just a handful of examples, the way a person can learn a new word after hearing it once or twice.
- instruction tuninginstruction tuning, n.: Training a language model on many tasks written as plain instructions with correct answers, so it learns to do what a request asks instead of just continuing the text.
- transfer learningtransfer learning, n.: Reusing what a model learned on one task to get a head start on another — like a pianist picking up the organ faster than a beginner. It lets models learn new jobs from little data.
- CLIPCLIP, prop. n.: An OpenAI model that learned to match pictures with their captions from hundreds of millions of web examples, so it can judge how well any piece of text describes any image.
- generalizationgeneralization, n.: How well a model copes with new examples it has never seen, rather than just memorizing its training data, like a student who understands the method, not just the answers.
- few-shot promptingfew-shot prompting, n.: Showing an AI a handful of examples of a task — inputs with the answers you want — inside the prompt, so it can copy the pattern for a new case.
- fine-tuningfine-tuning, n.: Taking an AI model that has already learned a lot and training it a little more on a smaller, focused set of examples so it gets better at one job or behaves a certain way.
Sources
- Palatucci, M., Pomerleau, D., Hinton, G. E. & Mitchell, T. M. (2009). Zero-Shot Learning with Semantic Output Codes. NIPS 2009.
- Wei, J. et al. (2022). Finetuned Language Models Are Zero-Shot Learners. ICLR 2022. arxiv.org/abs/2109.01652
Mentioned in
few-shot learningfew-shot learning, n.: Getting an AI system to handle a new task or recognize a new kind of thing from just a handful of examples, the way a person can learn a new word after hearing it once or twice., CLIPCLIP, prop. n.: An OpenAI model that learned to match pictures with their captions from hundreds of millions of web examples, so it can judge how well any piece of text describes any image., few-shot promptingfew-shot prompting, n.: Showing an AI a handful of examples of a task — inputs with the answers you want — inside the prompt, so it can copy the pattern for a new case., in-context learningin-context learning, n.: A language model's ability to pick up a new task from instructions or examples placed in its input, on the spot, without any retraining — like a quick study learning a game's rules from one round., instruction tuninginstruction tuning, n.: Training a language model on many tasks written as plain instructions with correct answers, so it learns to do what a request asks instead of just continuing the text., instruction-tuned modelinstruction-tuned model, n.: A language model given extra training on examples of instructions paired with good responses, so it does what you ask instead of merely continuing your text., MMLUMMLU, prop. n.: A giant multiple-choice exam for AI covering 57 subjects, from school math and history to law and medicine; for years it was the most quoted test of a model's general knowledge., one-shotone-shot, v.: Slang for an AI completing a whole task correctly on the first try from a single request — 'it one-shotted the app' — with no corrections or follow-up needed., text classificationtext classification, n.: Sorting pieces of text into categories automatically — what topic an article covers, whether a message is spam, which team should handle a customer's complaint., vision-language modelvision-language model, n.: An AI model that understands images and text together — it can describe a photo, answer questions about a chart or find the picture that matches a sentence., natural language inferencenatural language inference, n.: Deciding whether one sentence follows from another, contradicts it, or neither: if 'a man is playing a guitar on stage', is it true that 'a man is performing music'?, open-vocabulary detectionopen-vocabulary detection, n.: Object detection where you type what to look for, such as 'red umbrella' or 'forklift', and the system finds it, even if it was never trained on that exact category., time series foundation modeltime series foundation model, n.: A large AI model pretrained on huge numbers of time series, such as sales, traffic and sensor readings, that can forecast a new series it has never seen without being trained on it.