PulseAugur
EN
LIVE 09:48:23

New taxonomy improves LLM detection of coded language on social media · 2 sources tracked

By PulseAugur Editorial · [2 sources] ·

Researchers have developed a new taxonomy for identifying indirect linguistic expressions (ILE) used on social media platforms like TikTok and Bluesky to evade moderation. This taxonomy categorizes the underlying mechanisms of coded language rather than just the communicative intent. When integrated into LLM prompts, the proposed taxonomy significantly outperformed existing methods, improving accuracy by 4.7% and F1 score by 5.4% in detecting disguised meanings. AI

IMPACT This research could lead to more effective content moderation tools for social media platforms by improving LLMs' ability to detect disguised language.

RANK_REASON The cluster contains an academic paper detailing a new taxonomy for LLM-based coded language detection.

Read on arXiv cs.CL →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

New taxonomy improves LLM detection of coded language on social media · 2 sources tracked

COVERAGE [2]

  1. arXiv cs.CL TIER_1 English(EN) · Hamid Reza Firoozfar, Mohammadsadegh Abolhasani, Reza Mousavi, Paul Jen-Hwa Hu ·

    Beyond Surface Forms: A Comprehensive, Mechanism-Oriented Taxonomy of Indirect Linguistic Encoding for LLM-Based Coded Language Detection

    arXiv:2606.27314v1 Announce Type: new Abstract: To avoid moderation and surveillance on social media, some users routinely invent indirect linguistic expressions (ILE) that camouflage sensitive meanings. Such expressions surface as algospeak, euphemisms, and adversarial obfuscati…

  2. arXiv cs.CL TIER_1 English(EN) · Paul Jen-Hwa Hu ·

    Beyond Surface Forms: A Comprehensive, Mechanism-Oriented Taxonomy of Indirect Linguistic Encoding for LLM-Based Coded Language Detection

    To avoid moderation and surveillance on social media, some users routinely invent indirect linguistic expressions (ILE) that camouflage sensitive meanings. Such expressions surface as algospeak, euphemisms, and adversarial obfuscation, depending on intent and context, and they in…

RELATED ENTITIES

RELATED TOPICS