AI ethics evidence index Search
Source-linked daily record
Evidence record 18243 · automatically gathered

Multimodal Model Diffing for Feature Discovery and Control

Multimodal Large Language Models (MLLMs) exhibit strong visual understanding, yet the internal features that cause these behaviors remain difficult to identify, audit, or control. While applicable to post-hoc inspection, hidden states that are decomposed into interpretable feature directions using sparse autoencoders (SAEs) neither readily isolate which features are changed by multimodal training, nor are they directly useful for targeted control. We introduce MMDiff, a multimodal model-diffing

Record details

Published: 10 August 2026
Source: arXiv cs.AI
Category: Research
Topics: Transparency
Retrieved: 11 August 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: Transparency

KGCaRe: Explainable Complex Conditional Question Answering using Automatic Knowledge Graph Construction and Context Retrieval with LLMs

arXiv cs.AI · 10 August 2026

Shared topic: Transparency

Context and Symmetry in Auditing: A Case Study of Skeleton Inference in Motion Capture

arXiv · 10 August 2026

Shared topic: Transparency

Comprendia: AI-Augmented Code Comprehension

arXiv cs.HC · 10 August 2026

Shared topic: Transparency

Neuroevolution Arena: Nested Ecological Evaluation of Update-and-Inheritance Regimes across Neural Architectures

arXiv · 10 August 2026

Shared topic: Transparency

Visual-to-Haptic Augmentation in XR: A Wearable Glove for Perceptual Grounding in Multimodal Interaction

arXiv cs.HC · 11 August 2026

Shared topic: Transparency

Is the ACL Responsible NLP Checklist a Box-Ticking Exercise? A Large-Scale Analysis of EMNLP 2025

arXiv cs.CL (ethics-relevant NLP) · 10 August 2026

How to cite this record

ethics.ai (10 August 2026), “Multimodal Model Diffing for Feature Discovery and Control,” evidence record 18243, https://ethics.ai/record/18243 (originally published by arXiv cs.AI).

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.