reverse‑SynthID: Open‑Source Tool for Detecting and Removing Google Gemini’s Invisible SynthID Watermark

reverse‑SynthID is an open‑source Python project that uses FFT‑based spectral analysis and multi‑resolution codebooks to detect Google Gemini’s invisible SynthID watermark with about 90% accuracy and to remove it, achieving up to 43 dB PSNR and a 91% drop in phase coherence.

Google GeminiPythonSynthID

0 likes · 12 min read

reverse‑SynthID: Open‑Source Tool for Detecting and Removing Google Gemini’s Invisible SynthID Watermark

AntTech

May 15, 2025 · Artificial Intelligence

Live Deep Dive into Two Award‑Winning WSDM 2025 Papers on Popularity Bias in Recommendation Models and Graph‑Based Causal Inference

This announcement introduces a live session that will dissect two best‑paper award research works from WSDM 2025—one revealing how recommendation models amplify popularity bias through spectral analysis and proposing a lightweight regularizer, and the other presenting a graph disentangle causal model that integrates GNNs with structural causal models to improve causal inference on networked observational data.

Recommendation SystemsWSDM 2025causal inference

0 likes · 4 min read

Live Deep Dive into Two Award‑Winning WSDM 2025 Papers on Popularity Bias in Recommendation Models and Graph‑Based Causal Inference

AntTech

May 8, 2025 · Artificial Intelligence

How Recommendation Models Amplify Popularity Bias: A Spectral Perspective and the ReSN Correction Method

The award‑winning WSDM 2025 paper reveals how recommendation models magnify popularity bias through spectral alignment of the rating matrix’s leading singular vector with item popularity, and proposes the ReSN regularization technique to mitigate this bias efficiently.

FairnessRecommendation SystemsRegularization

0 likes · 8 min read

How Recommendation Models Amplify Popularity Bias: A Spectral Perspective and the ReSN Correction Method