Instruction Tuning Vol. 2

·Sebastian Ruder··

Last month, we covered the first generation of instruction-tuning datasets that have been mainly based on existing NLP tasks.This month, we cover the latest datasets that are now much closer to real-world use cases—but still have their limitations! Initiatives to Get Involved in AI ResearchBefore we get into these, here are a few initiatives to get involved in AI research:The Aya Project: Help build state-of-the-art multilingual LLMs and make LLMs accessible in your language. Aya launched in Jan...

Read full article →

Related Articles

Accelerating Gemma 4: faster inference with multi-token prediction drafters
amrrs · Hacker News · 4d ago
OpenAI’s o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors
donsupreme · Hacker News · 6d ago
ProgramBench: Can language models rebuild programs from scratch?
jonbaer · Hacker News · 2d ago
ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters
steveharing1 · Hacker News · 2d ago
A couple million lines of Haskell: Production engineering at Mercury
unignorant · Hacker News · 6d ago