Showing cs.CLShow all
2 papers · 1 filter
cs.CL2026
WavePrune: One period is often enough for RoPE
Guancheng Du, Luotian Huang, Shaowen Wang +2
Rotary Position Embedding (RoPE) encodes token positions by rotating each two-dimensional channel of the query and key vectors at a channel-specific frequency, making the attention…
cs.CL2025
SagaScale: A Realistic, Scalable, and High-Quality Long-Context Benchmark Built from Full-Length Novels
Guancheng Du, Yong Hu, Wenqing Wang +2
Large Language Models (LLMs) have shown significant progress, but understanding long and complex documents remains challenging. Many long-context benchmarks have been proposed, but…