Research map / Large language models / Language, safety and interpretability
Defending Against Neural Fake News
Rowan Zellers, Ari Holtzman, Hannah Rashkin, Yonatan Bisk, Ali Farhadi, Franziska Roesner, Yejin Choi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use metadata as context when generating article bodies | GROVER · April 2019 test set | perplexity: between .6 to .9 perplexity points | Improved |
| Increase model size to GROVER-Mega | GROVER · April 2019 test set, conditional | perplexity: 8.7 | Improved |
| Use GROVER as a discriminator rather than the best existing discriminators | Neural fake news detection · Unpaired | accuracy: 73% -> 92% | Improved |
| Rewrite propaganda articles with GROVER | Human Propaganda · Propaganda articles | overall trustworthiness: 2.19 to 2.42 | Improved |
| Use a larger generator than the discriminator | GROVER detection · Neural fake news detection | accuracy: below 81% | Hurt |
| Use a larger discriminator than the generator | GROVER detection · Neural fake news detection | accuracy: above 98% | Improved |
| Add weakly supervised examples generated by GROVER-Large when few GROVER-Mega examples are available | GROVER discriminator · Detection of GROVER-Mega generations | accuracy: around 50% -> around 78% | Improved |
| Use random sampling without variance restriction | GROVER generation · Generated article bodies | perplexity | Hurt |
| Use nucleus sampling to restrict generation variance | GROVER generation · Generated article bodies | perplexity | Improved |
| Use an excessively low top-p threshold | GROVER generation · Neural fake news detection | discrimination difficulty | Hurt |
Closest papers by meaning
- CoT is Not the Chain of Truth: An Empirical Internal Analysis of Reasoning LLMs for Fake News Generation (ICML 2026 · similarity 0.78)
- Contradicted in Reliable, Replicated in Unreliable: Dual-Source Reference for Fake News Early Detection (AAAI 2025 · similarity 0.78)
- Harnessing Network Effect for Fake News Mitigation: Selecting Debunkers via Self-Imitation Learning (AAAI 2024 · similarity 0.78)
- Bad Actor, Good Advisor: Exploring the Role of Large Language Models in Fake News Detection (AAAI 2024 · similarity 0.78)
- Drifting Away from Truth: GenAI-Driven News Diversity Challenges LVLM-Based Misinformation Detection (AAAI 2026 · similarity 0.76)
- Is LLMs Hallucination Usable? LLM-based Negative Reasoning for Fake News Detection (AAAI 2025 · similarity 0.75)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
