Advances in AI-Generated Text Detection and Decentralized AI Platforms

The field of natural language processing is witnessing significant developments in the detection of AI-generated text and the deployment of decentralized AI platforms. Researchers are exploring innovative approaches to detect and mitigate the risks associated with AI-generated text, including the use of adversarial training and perturbation-invariant feature engineering. Meanwhile, decentralized AI platforms are being proposed to address the trust and cost issues associated with centralized LLM services. These platforms leverage blockchain technology and crowdsourcing to enable secure and efficient model deployment and inference. Noteworthy papers in this area include Modeling the Attack, which presents a novel detection framework that achieves state-of-the-art performance in detecting AI-generated text, and PolyLink, which introduces a blockchain-based decentralized AI platform for LLM inference. Other notable works include EditLens, which proposes a regression model to quantify the extent of AI editing in text, and Audit the Whisper, which introduces a calibrated auditing pipeline to detect steganographic collusion in multi-agent LLMs.

Sources

Modeling the Attack: Detecting AI-Generated Text by Quantifying Adversarial Perturbations

PolyLink: A Blockchain Based Decentralized Edge AI Platform for LLM Inference

EditLens: Quantifying the Extent of AI Editing in Text

Time Is Effort: Estimating Human Post-Editing Time for Grammar Error Correction Tool Evaluation

Audit the Whisper: Detecting Steganographic Collusion in Multi-Agent LLMs

Protecting De-identified Documents from Search-based Linkage Attacks

A Formal Framework for Fluency-based Multi-Reference Evaluation in Grammatical Error Correction

Overview of the Plagiarism Detection Task at PAN 2025

OBJVanish: Physically Realizable Text-to-3D Adv. Generation of LiDAR-Invisible Objects

Machines in the Crowd? Measuring the Footprint of Machine-Generated Text on Reddit

Built with on top of