WnW: Waxing-and-Waning KV Cache for Long-Form Speech LLMs

Shared on Bluesky by 2 AI experts

• A arxiv cs.CL @arxiv-cs-cl.bsky.social: Yiming Yao, Chenyang Lyu, Xuanfan Ni, Longyue Wang, Weihua Luo, Yazheng Yang, Jinsong Su WnW: Waxing-and-Waning KV Cache for Long-Form Speec… →
• A AI Firehose @ai-firehose.column.social: Researchers introduce WnW (Waxing-and-Waning KV cache), a new method that reduces memory usage for long-form speech LLMs, preserving near-fu… →