Recurrent Language Model Closes Long-Context Gap Without Extra Memory in Princeton Research
· techtimes.com · popularity 19/100 · 0 similar stories today · AI & Data Science (Artificial Intelligence)
Read the original at techtimes.com →
Summary
Recurrent language model long-context accuracy just got cheaper to improve: Princeton researchers introduced MERLIN, a plug-in routing layer that tracks key-variance in Multi-State Linear Attention models and gains 8.3 percentage points on needle-in-a-haystack retrieval, beating models that consume four times more memory at no additional inference cost.
Found in: Tech Times. Rating basis: single outlet; outlet tier 2 (unlisted); 1 feeds; feed position 7.