Search papers, labs, and topics across Lattice.
5
0
9
0
Classic GNNs still dominate misinformation detection, achieving >20% F1 gains over MLPs with comparable or better inference times, challenging the trend towards ever-larger models.
Clickbait detection can be both accurate and lightning-fast: graph-based models rival slower, more complex methods while slashing inference time.
LLMs consistently mark down grammar and conventions more harshly than human graders, revealing a systematic bias that can be detected with surprisingly little data.
Half of news articles contain reused sentences, and you'd miss it if you only looked for exact copies.
Fine-tuning AI-generated text detectors on a per-LLM or per-LLM-family basis skyrockets detection accuracy to nearly 100%, leaving generic detectors in the dust.
