A short rant on the data gap that keeps slowing down Persian language model research.
Placeholder post — replace with real write-up.
Notes to expand on:
- Which tasks are most starved for labeled Persian data
- What's already out there and where it falls short
- Whether a small, well-curated dataset beats a big noisy one for this