A new small language model named Aurora-80K has been released, featuring 80,000 parameters and a 4,096-token vocabulary. Initial benchmarks show a Wikitext-2 BPB score of 3.2902, a BLiMP score of 52.31%, and an Arc-Easy score of 26.05%. Further details are available on its Hugging Face model page. AI
IMPACT Provides a new, small-scale model option for researchers and developers experimenting with language models.
RANK_REASON Release of a small language model with performance benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
