OpenEuroLLM / OpenEuroLLM/Taskboard

T4.3 Compare model accuracy with 1 vs 2 high resource languages

Open
#135 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
No language data
Stars
3
Forks
0
PR merge metrics
No merged PRs in 30d

Description

Our basic understanding is that we should overweight English because it has the most high quality data, but there is also high quality data in other larger languages. If we could get roughly equal quality english and german samples, for example, it would help answer a question that I think is worth investigating:

  • Train model with 80/20 english + low resource language
  • Train model with 80/20 german(?) + low resource language
  • Train model with 40/40/20 english + german(?) + low resource language

My intuition is that 80/20 will do better than 40/40/20 even with roughly equivalent quality data, but it would be good to confirm it.

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No files, tests, or entry points are identified in the issue. Define the English, German, and low-resource datasets and training setup, then compare accuracy for the proposed 80/20, 80/20, and 40/40/20 mixtures; done means reporting the results against the stated hypothesis.

Written by the indexing model from the issue text.

Assessment

Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.