RCStat: A Statistical Framework for using Relative Contextualization in Transformers

arXiv:2506.19549 · cs.CL, cs.AI, cs.LG · Submitted 2025-06-24 · Read on arXiv

cs.CL, cs.AI, cs.LG

Submitted: 2025-06-24

Updated: 2026-10-13

Code: https://github.com/NVIDIA/kvpress

Terminology

• Relative Contextualization — 2× in this paper
• Transformer — 2× in this paper · explained in 2 episodes
• Alignment — 1× in this paper · explained in 1 episode
• Benchmark — 1× in this paper · explained in 2 episodes
• Fidelity — 1× in this paper · explained in 2 episodes

Sources

• Quantifying Attention Flow in Transformers
• Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation
• PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling
• Palu: Compressing KV-Cache with Low-Rank Projection
• Identifying Linear Relational Concepts in Large Language Models
• Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps
• Learning to Attribute with Attention
• A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression
• Get More with LESS: Synthesizing Recurrence with KV Cache Compression for Efficient LLM Inference
• How GPT learns layer by layer
• Transcoders Find Interpretable LLM Feature Circuits
• Trapping LLM Hallucinations Using Tagged Context Prompts
• Ada-KV: Optimizing KV Cache Eviction by Adaptive Budget Allocation for Efficient LLM Inference
• Q-Filters: Leveraging QK Geometry for Efficient KV Cache Compression
• The Llama 3 Herd of Models
• When Attention Sink Emerges in Language Models: An Empirical View
• A Survey on Large Language Model Acceleration based on KV Cache Management
• MatryoshkaKV: Adaptive KV Compression via Trainable Orthogonal Projection
• Copy Suppression: Comprehensively Understanding an Attention Head
• Using Captum to Explain Generative Language Models

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering