Showing cs.CLShow all
3 papers · 1 filter
cs.CL2026
Compression Does Not End at Selection: Post-Selection Positional Geometry for Sparse KV Caches
Wanyang Hong, Zhaoning Zhang
Sparse KV compression reduces long-context inference costs by retaining selected historical states, yet these states typically keep their coordinates from the full history. Under R…
cs.CL2026
Role-aware Heuristic Episodic Attention for Conversational LLMs
Wanyang Hong, Zhaoning Zhang, Yi Chen +5
Large language models often lose track of persistent instructions and relevant information as multi-turn conversations grow. We study this cumulative contextual decay through three…
cs.CL2025
Rhea: Role-aware Heuristic Episodic Attention for Conversational LLMs
Wanyang Hong, Zhaoning Zhang, Yi Chen +5
Large Language Models (LLMs) have achieved remarkable performance on single-turn tasks, yet their effectiveness deteriorates in multi-turn conversations. We define this phenomenon…