Home › Hard problems › Inside the model

Inside the model: the core of LLMs

Learning from far less data

Solo: possibleOne person can test a narrow slice of it.

a child:about 100 million wordsa model:over 10 trillion words(this tower would gofar off the page)same language, about100,000 times the readingHow does a child do it?
A child learns language from about 100 million words; models need over 10 trillion, and nobody knows how to close most of that gap.

Why it is still open

A child learns language from about 100 million words; models need over 10 trillion. Nobody knows how to close most of that gap.

An experiment one person can run in 8 weeks

The BabyLM challenge trains models on 100 million words or fewer. Enter it with one idea (curriculum order, a data quality filter).

Verdict: Solo: possible

BabyLM exists so that small teams can compete.

The solo test behind the verdict

  1. Can it be measured? There is a number that shows you did better, such as an error rate, a catch rate or a cost per answer.
  2. Is it cheap to try? It runs on a laptop, free open models and open data, or under about ₹50,000 of rented GPU time.
  3. Is there a public yardstick? A benchmark or dataset others use, so your result can be compared and believed.
  4. Would a small win get noticed? A lab, a funder or a customer would care about a 10% improvement, not only a full solution.

Four yeses means Solo: strong. Two or three means Solo: possible. Fewer means Follow, don't build.

PreviousNew architectures beyond the transformer

All 79 hard problems · Working on this, or know of a result we missed? basant@cleandataengineering.ai