Investigate mapping token embeddings from source to target
Nobody has claimed this yet.
- Dominant language
- Python
- Stars
- 39
- Forks
- 7
- Avg merge
- 1d 9h
- Merged PRs (30d)
- 5
Description
A recently published paper introduced a strategy called "trans-tokenization", which "focuses on adapting a high-resource monolingual LLM to an unseen target language by initializing the token embeddings of the target language using a weighted average of semantically similar token embeddings from the source language." We should investigate whether this approach could improve the performance of adding trained tokens to NLLB.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by reading the linked trans-tokenization paper and reviewing how trained tokens are currently added to NLLB. Determine whether weighted initialization from semantically similar source-language embeddings could improve performance, and document the evaluation needed to support the conclusion.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100