TraceKV: Policy-Preserving Low-Rank KV Cache Compression for Reasoning Vision-Language Models
cs.CV, cs.AI, cs.LG
Submitted: 2026-11-11
Updated: 2026-11-11
Code: https://github.com/AK-DREAM/TraceKV
Terminology
Sources
- Qwen3-VL Technical Report
- Qwen2.5-VL Technical Report
- Distilling the Knowledge in a Neural Network
- DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
- BELT: BEhavior-Aligned Low-Rank KV Cache Compression with Learned Top-k Rank Allocation
- LSVD: Loss-Aware Low-Rank Approximation for Efficient Low-Precision Vision-Language Models
- ASVD: Activation-aware Singular Value Decomposition for Compressing Large Language Models
- iS-KV: Online Low-Rank KV Cache Compression via Block-Incremental SVD
- InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs