Timur Seidalin

Tim2190's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in stukenov/sozkz-morphbpe-256k-kk-v1 4 months ago

Independent fertility benchmark — sozkz-morphbpe-256k is the least-fragmenting tokenizer for Kazakh (5-way comparison)

#1 opened 4 months ago by

Tim2190

New activity in Tim2190/kaz-rag-search-benchmark 4 months ago

Morphology beats multilingual embeddings for Kazakh retrieval — a 300-query benchmark

#2 opened 4 months ago by

Tim2190

New activity in ibm-granite/granite-embedding-278m-multilingual 4 months ago

Independent Kazakh (kk) retrieval eval of granite-embedding-278m — strong on morphology, sharp drop on synonym queries

#5 opened 4 months ago by

Tim2190