base model
In plain wordsA language model fresh from its first broad training on text, before being taught to follow instructions or chat. It continues whatever text it is given instead of answering.
- 1
A language modellanguage model, n.: A system that has learned how likely different sequences of words are, so it can predict the next word in a text and, by doing that repeatedly, write text of its own. that has completed 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. (and possibly mid-trainingmid-training, n.: A middle stage in building a language model, between its broad initial training and its tuning into an assistant: it keeps learning the same way, but from a smaller, carefully chosen diet of math, code, reasoning or long documents.) but no instruction or preference post-trainingpost-training, n.: Everything done to an AI model after its initial bulk training on huge amounts of text: the extra teaching that turns a raw text predictor into a helpful, safe assistant or a careful reasoner.. A base model is a pure next-token predictor: given a 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., it continues the text in whatever genre the context implies, so a question may be followed by more questions rather than an answer. Using one well means framing the task as a document to be completed, often with few-shot examplesfew-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.. Base models are released for research and as starting points for 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., and are valued for more diverse samples and better-calibratedcalibration, n.: How well a model's stated confidence matches reality: when a well-calibrated system says it is 80% sure, it turns out to be right about 80% of the time. probabilities, both of which post-training tends to reduce. They have no safety training of their own.
The base model just kept writing quiz questions, so we added three worked examples to the prompt.
- 2
Fine-tuning
More loosely, whatever checkpoint a 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. run starts from, even one that is already instruction-tuned, as in 'we used an 8B instruct model as the base model for our adapter'. Context usually shows which sense is meant.
Origin
'Base model' has long meant whatever model further training starts from, so its earliest use in the pretrained-only sense is not established. That sense was standard by early 2023, when OpenAI's fine-tuning documentation offered only the 'base models' davinci, curie, babbage and ada — its original GPT-3 models, without the instruction training of its InstructGPTInstructGPT, prop. n.: A 2022 OpenAI language model trained with human feedback to follow instructions; people preferred its answers to those of a model 100 times larger, and its recipe led directly to ChatGPT.-derived models. Open releases such as Meta's Llama 2 (July 2023) then made paired base and chat checkpoints the norm.
Usage note
Overlaps with but differs from foundation modelfoundation model, n.: A very large AI model, built once at great cost from vast amounts of text, images or code gathered mostly from the internet, that many products then reuse as their starting point, from chatbots to coding tools.: a foundation model may be a base model or a post-trained one. 'Pretrained model' is the older synonym from the BERT era, when nearly every released model was a base model.
Related
- 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.
- 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.
- chat modelchat model, n.: A language model trained to hold a back-and-forth conversation, keeping track of who said what and replying as a helpful assistant. It is the engine inside chat apps.
- post-trainingpost-training, n.: Everything done to an AI model after its initial bulk training on huge amounts of text: the extra teaching that turns a raw text predictor into a helpful, safe assistant or a careful reasoner.
- 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.
- next-token predictionnext-token prediction, n.: The simple task most AI chatbots were trained on: given some text, guess the next small piece of it. Repeating that guess, piece by piece, produces whole answers.
- 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.
- chat modelchat model, n.: A language model trained to hold a back-and-forth conversation, keeping track of who said what and replying as a helpful assistant. It is the engine inside chat apps.
Sources
- Ouyang, L. et al. (2022). Training language models to follow instructions with human feedback. NeurIPS 2022. arxiv.org/abs/2203.02155
Mentioned in
large language modellarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code., post-trainingpost-training, n.: Everything done to an AI model after its initial bulk training on huge amounts of text: the extra teaching that turns a raw text predictor into a helpful, safe assistant or a careful reasoner., 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., chat modelchat model, n.: A language model trained to hold a back-and-forth conversation, keeping track of who said what and replying as a helpful assistant. It is the engine inside chat apps., fancy autocompletefancy autocomplete, n.: A dismissive description of AI chatbots as nothing more than a souped-up version of the feature on a phone keyboard that suggests your next word., 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., foundation modelfoundation model, n.: A very large AI model, built once at great cost from vast amounts of text, images or code gathered mostly from the internet, that many products then reuse as their starting point, from chatbots to coding tools., instruction datasetinstruction dataset, n.: A collection of example requests paired with good answers, used to teach a pretrained language model to follow instructions and act as a helpful assistant., 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., mid-trainingmid-training, n.: A middle stage in building a language model, between its broad initial training and its tuning into an assistant: it keeps learning the same way, but from a smaller, carefully chosen diet of math, code, reasoning or long documents., stop sequencestop sequence, n.: A piece of text that tells an AI model to stop writing as soon as it produces it — for example, stopping at a blank line or just before the next question., crosscodercrosscoder, n.: A concept-finding tool that looks at several layers, or several versions of a model, at once, so that one shared concept can be tracked wherever it appears., model diffingmodel diffing, n.: Comparing the insides of two closely related AI models — such as a model before and after extra training — to find exactly which internal concepts and mechanisms changed., simulator theorysimulator theory, n.: A way of thinking about AI language models as simulators that can play any character found in their training text, rather than as a single agent with its own beliefs and goals., superficial alignment hypothesissuperficial alignment hypothesis, n.: The idea that a language model learns nearly all its knowledge and skills from its initial bulk training, and that later tuning mainly teaches it the style and format to use when talking to people.