LoopKV: Loop-Aware KV Cache Compression for Serving Looped Transformers
cs.LG, cs.CL, cs.DC
Submitted: 2026-11-03
Updated: 2026-11-03
Comments: 28 pages, 15 figures
Project page: https://nvidia.github.io/Model-Optimizer/announcements/20261008-autoquant-kv.html
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation
- Read What Matters: Query-Adaptive Quantization for KV Caches
- A Mechanistic Analysis of Looped Reasoning Language Models
- xKV: Cross-Layer KV-Cache Compression via Aligned Singular Vector Extraction
- Training Verifiers to Solve Math Word Problems
- Memory-Efficient Looped Transformer: Decoupling Compute from Memory in Looped Language Models
- WaveFront Decoding: Parallelized Self-Speculative Decoding for Looped Language Models
- ResidualQuant: KV Cache Quantization for Looped Transformers with 2-Bit Residuals
- The Surprising Effectiveness of Shared Memory in Looped Transformers
- Looped Latent Attention: Cross-Loop KV Compression for Looped Transformers
- SwiftKV: Fast Prefill-Optimized Inference with Knowledge-Preserving Model Transformation
- Microscaling Data Formats for Deep Learning
- AlignedKV: Reducing Memory Access of KV-Cache with Precision-Aligned Quantization
- Parallel Loop Transformer for Efficient Test-Time Computation Scaling
- No Token Left Behind: Reliable KV Cache Compression via Importance-Aware Mixed Precision Quantization
- FlashLoop: Fast and Memory-Efficient Looped Transformers via Lazy Updates
- KVSharer: Efficient Inference via Layer-Wise Dissimilar KV Cache Sharing
- TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate
- RDKV: Rate-Distortion Bit Allocation for Joint Eviction and Quantization of the KV Cache
- OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis