Adaptive Filtering of the KV Cache: Diagnosing and Correcting Structural-Role Bias in LLM Inference
Attention-based KV cache eviction (H2O and its descendants) compresses the memory-constrained state of a long-context model by ranking tokens on accumulated attention mass, treated here as signal energy, and keeping the heaviest. On schema-dense input streams such as nested JSON, this score acts as a non-stationary filter that disproportionately retains noise: a non-content sink role (delimiters or whitespace) carries an order of magnitude more energy than any content role, and structural KEY to
Record details
Published: 14 July 2026
Source: arXiv
Category: Research
Topics: Bias & fairness · Healthcare · Environment
Retrieved: 16 July 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
An empirical evaluation of the risks of AI model updates using clinical data: stability, arbitrariness, and fairness
arXiv · 27 April 2026
Multi-omics approaches for understanding gene-environment interactions in noncommunicable diseases: techniques, translation, and equity issues
OpenAlex · 31 January 2025
Price of Fairness in Bandits: A Tight Minimax Characterization
arXiv · 15 July 2026
LoRA-Based Cascaded Multimodal Fusion for Action Recognition in Medical Training Environments
arXiv cs.AI · 13 July 2026
Developing a Core Outcome Set for the Evaluation of Remote Patient Monitoring Interventions Using the Sextuple Aim: Modified Delphi Study
JMIR (Journal of Medical Internet Research) · 15 July 2026
Auditing Fairness-Privacy Trade-offs: Subpopulation-Level Effects of Fairness-Enhancing Algorithms
arXiv · 16 July 2026
How to cite this record
ethics.ai (14 July 2026), “Adaptive Filtering of the KV Cache: Diagnosing and Correcting Structural-Role Bias in LLM Inference,” evidence record 10610, https://ethics.ai/record/10610 (originally published by arXiv).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.