LLM Digest

Story

arxiv_cs_cl · Nov 3, 2026 · paper

Source brief

What Does the Model See? Representation Validity in Multimodal LLM-Based Video Analysis

arxiv.orgNov 3, 2026
original source linked

In brief

Multimodal large language models (MLLMs) allow social scientists to analyze video at scale, but researchers must decide which parts of a video to provide as input. We use the term representation validity to examine ho...

Feed lens

eval

Read the original at arxiv.org →Open in live feed Read that day’s brief

What Does the Model See? Representation Validity in Multimodal LLM-Based Video Analysis

Earlier in this thread 4 items

LSEAD: A Privacy-Preserving LLM-Based Speech Analysis Framework for Early Alzheimer's Disease Screening

SpaCellAgent: A Self-Evolving LLM-Based Multi-Agent Framework for Trajectory Analysis

Symmetric Divergence and Normalized Similarity: A Unified Topological Framework for Representation Analysis

Specificity Bias and the Wikipedia Paradox: An xAI Analysis of AI Text Detectors under Domain Shift