AI ethics evidence index Search
Source-linked daily record
Evidence record 14957 · automatically gathered

A Cross-Architecture Audit of Direction-Based Inference-Time Defences in Vision-Language Models

Inference time defences against vision language model jailbreaks often subtract a calibrated direction from the residual stream at a chosen decoder layer. We compare five defence candidates across 15 model and layer cells from four architectural families under a magnitude controlled protocol that matches the intervention size for each prompt and pairs every direction with a random control of the same norm. The candidates are the mean image conditioning shift, a CMRM style refusal direction, a Sh

Record details

Published: 30 July 2026
Source: arXiv
Category: Research
Topics: Safety & alignment · Military & security · Transparency
Retrieved: 31 July 2026

source-onlyevidence status

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.

Shared topic: Safety & alignment · Shared topic: Military & security

Acceptance Cards:A Four-Diagnostic Standard for Safe Fine-Tuning Defense Claims

arXiv · 11 May 2026

Shared topic: Safety & alignment · Shared topic: Military & security

Defending Against Backdoor Attacks via Alignment Checking in Model-Contrastive Federated Learning

arXiv · 29 July 2026

Shared topic: Military & security · Shared topic: Transparency

A First Look at Coding Agents' Compliance with AI Contribution Rules in Open-Source Communities

arXiv cs.AI · 29 July 2026

Shared topic: Safety & alignment · Shared topic: Military & security

Recover, Decode, Reguard: Guard-Agnostic Defense Amplification againstEncoded VLM Jailbreaks

arXiv red teaming query · 29 July 2026

Shared topic: Safety & alignment · Shared topic: Transparency

Auditing Alignment Controllability in LLMs via Political Axes

arXiv cs.CY · 28 July 2026

Shared topic: Safety & alignment · Shared topic: Transparency

KANEx: Translating Kolmogorov-Arnold Networks' Interpretability to Medical Explainability

arXiv cs.AI · 27 July 2026

How to cite this record

ethics.ai (30 July 2026), “A Cross-Architecture Audit of Direction-Based Inference-Time Defences in Vision-Language Models,” evidence record 14957, https://ethics.ai/record/14957 (originally published by arXiv).

JSON

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.