Zhao et al.
Low-rank KV cache compression preserves all tokens during long reasoning
iS-KV uses block-incremental SVD to update both basis and coefficients, avoiding representation drift and matching original model accuracy at 4-5x compression on DeepSeek-R1 and Qwen3-8B.
Research Lab
Yiren Zhao · Guanghui Song · Tianrui Qin · Kejiang Ye · Cheng-zhong Xu · Xitong Gao
The Hong Kong University of Science and Technology (Guangzhou) · Guangdong OPPO Mobile Telecommunications Corp., Ltd. · Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · University of Macau · Shenzhen University of Advanced Technology
Research Digest··3 min read
The authors propose iS-KV, an online low-rank KV-cache compression method for long chain-of-thought reasoning.
Why this paper
From Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences and 4 others
In one line
iS-KV retains all KV-cache positions through synchronized online low-rank compression, preserving reasoning accuracy better than token eviction at matched memory budgets.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks (2 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§