Cognitive Inception: Agentic Reasoning against Visual Deceptions by Injecting Skepticism

TL;DR

This paper introduces Inception, a reasoning framework that enhances large language models' ability to detect visual deceptions by injecting skepticism, inspired by human cognition, significantly improving authenticity verification against AI-generated visual content.

Contribution

The paper presents the first fully reasoning-based framework, Inception, that uses iterative skepticism to improve LLMs' visual deception detection capabilities.

Findings

Achieved significant performance improvements over existing LLM baselines.

Outperformed state-of-the-art methods on the AEGIS benchmark.

Demonstrated the effectiveness of skepticism injection in reasoning processes.

Abstract

As the development of AI-generated contents (AIGC), multi-modal Large Language Models (LLM) struggle to identify generated visual inputs from real ones. Such shortcoming causes vulnerability against visual deceptions, where the models are deceived by generated contents, and the reliability of reasoning processes is jeopardized. Therefore, facing rapidly emerging generative models and diverse data distribution, it is of vital importance to improve LLMs' generalizable reasoning to verify the authenticity of visual inputs against potential deceptions. Inspired by human cognitive processes, we discovered that LLMs exhibit tendency of over-trusting the visual inputs, while injecting skepticism could significantly improve the models visual cognitive capability against visual deceptions. Based on this discovery, we propose \textbf{Inception}, a fully reasoning-based agentic reasoning framework to conduct generalizable authenticity verification by injecting skepticism, where LLMs' reasoning logic is iteratively enhanced between External Skeptic and Internal Skeptic agents. To the best of our knowledge, this is the first fully reasoning-based framework against AIGC visual deceptions. Our approach achieved a large margin of performance improvement over the strongest existing LLM baselines and SOTA performance on AEGIS benchmark.