Skip to content

Explore everything

Filter by research area, year and venue. Every filter is in the address bar, so you can share a view as a link.

LLMs & generative AIPlatforms, policy & society Jailbroken: How Does LLM Safety Training Fail?This paper investigates why large language models trained for safety remain vulnerable to jailbreak attacks. Through analysis of models including GPT-4 and Claude v1.3, the authors identify two key failure modes: competing objectives (where safety constraints conflict with other capabilities) and mismatched generalization (where safety… Wei et al. · arXiv · 2023

LLMs & generative AIDetection methods A Survey on LLM-Generated Text Detection: Necessity, Methods, and Future DirectionsAs LLMs become ubiquitous, detecting machine-generated text is critical for mitigating risks of misinformation, academic dishonesty, and misuse. This survey comprehensively reviews detection strategies—watermarking techniques, statistical methods, neural-based detectors, and human-assisted approaches—along with datasets, evaluation… Wu et al. · arXiv · 2023

LLMs & generative AIDatasets & evaluation How Do Large Language Models Capture the Ever-changing World Knowledge? A Review of Recent AdvancesThis survey reviews methods for aligning large language models with ever-changing world knowledge without retraining from scratch. The authors systematically categorize approaches as either implicit (modifying LLM parameters directly via knowledge editing or continual learning) or explicit (augmenting LLMs with external resources via… Zhang et al. · arXiv · 2023

LLMs & generative AIPlatforms, policy & society Universal and Transferable Adversarial Attacks on Aligned Language ModelsDespite alignment training, modern language models can be induced to generate harmful content via automatically-discovered adversarial suffixes. The authors propose Greedy Coordinate Gradient (GCG) search, which generates a single universal suffix that transfers across multiple models and prompts with high success rates (88% on… Zou et al. · arXiv · 2023

LLMs & generative AIDetection methods Combating Misinformation in the Age of LLMs: Opportunities and ChallengesThis survey systematically reviews how large language models (LLMs) can both combat misinformation and generate it at scale. LLMs offer opportunities for detection, intervention, and attribution through their world knowledge and reasoning abilities, but also pose challenges as they can create deceptive content intentionally or through… Chen & Shu · arXiv · 2023

LLMs & generative AIDetection methods A Watermark for Large Language ModelsProposes a watermarking scheme for large language models that embeds imperceptible signals into generated text, enabling detection of machine-generated content without access to model parameters or APIs. The watermark works by promoting a randomized set of "green" tokens during decoding and can be detected using a statistical z-test; the… Kirchenbauer et al. · arXiv · 2023

LLMs & generative AIDetection methods DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability CurvatureThis paper identifies that text sampled from language models tends to occupy negative-curvature regions of the model's log-probability landscape, whereas human-written text occupies more neutral regions. DetectGPT leverages this observation to detect machine-generated text in a zero-shot manner by analyzing probability curvature, without… Mitchell et al. · 40th International Conference on Machine… · 2023

LLMs & generative AIDetection methods On the Risk of Misinformation Pollution with Large Language ModelsThis paper investigates how modern LLMs can be misused to generate convincing misinformation at scale and its impact on downstream NLP applications. Using GPT-3.5 as a misinformation generator, the authors demonstrate that synthetically-generated false information can significantly degrade ODQA system performance (up to 87% in some… Pan et al. · arXiv · 2023

LLMs & generative AIDetection methods Can AI-Generated Text be Reliably Detected?This paper stress-tests AI-text detectors using a novel recursive paraphrasing attack, showing that watermarking-based, neural network-based, and retrieval-based detectors can be reliably broken with only minor text degradation. The work establishes a theoretical connection between detector AUROC and the total variation distance between… Sadasivan et al. · arXiv · 2023

Detection methodsLLMs & generative AI 3HAN: A Deep Neural Network for Fake News DetectionProposes 3HAN, a three-level hierarchical attention network for fake news detection that models articles as a hierarchy of words, sentences, and headlines. The model assigns differential importance weights to different parts via attention mechanisms, achieving 96.77% accuracy while providing interpretable attention visualizations showing… Singhania et al. · arXiv · 2023

LLMs & generative AIDetection methods Disinformation Capabilities of Large Language ModelsComprehensive empirical evaluation of 10 LLMs' ability to generate disinformation news articles. Most models readily generate convincing disinformation that agrees with dangerous narratives; Falcon is the sole exception. Vykopal et al. · arXiv · 2023

Detection methodsLLMs & generative AI Linguistic-style-aware Neural Networks for Fake News DetectionHERO (Hierarchical Recursive Neural Network) detects fake news by learning the linguistic style encoded in hierarchical linguistic trees — structures that jointly capture syntactic constituency and rhetorical discourse organization from word to document. Unlike prior handcrafted approaches that extract node-level statistics (word… Zhou et al. · arXiv · 2023

LLMs & generative AIDetection methods All the News That's Fit to Fabricate: AI-Generated Text as a Tool of Media MisinformationThree experiments investigate whether AI-generated news text can be credible and influence public opinion. The authors find that GPT-2 models (355M to 1.5B parameters) generate text nearly indistinguishable from human-written baseline news, particularly when politically congruent stories are paired with partisan audiences. Kreps et al. · Journal of Experimental Political Science · 2022

Spread & networksDetection methods "This is Fake! Shared it by Mistake": Assessing the Intent of Fake News SpreadersThis work is the first to systematically assess whether individuals spread fake news intentionally or unintentionally. The authors propose an influence graph that captures psychological (internal) and social (external) factors driving unintentional spreading, validate it on two datasets with manually-annotated spreader intent, and… Zhou et al. · ACM Web Conference 2022 · 2022

Detection methodsDatasets & evaluation From Fake News to #FakeNews: Mining Direct and Indirect Relationships among Hashtags for Fake News DetectionHashtags reveal news credibility signals during propagation: fake news clusters with hashtags used primarily for misinformation, while true news uses hashtags associated with reliable content. The authors propose NewsTag, a language-independent semi-supervised algorithm that mines both direct co-occurrence and indirect relationships… Zhou et al. · ACM Conference · 2022

LLMs & generative AIDetection methods Discovering Latent Knowledge in Language Models Without SupervisionLanguage models trained with standard objectives (imitation, reward-based) can output false statements even when they internally "know" the truth. This paper proposes Contrast-Consistent Search (CCS), an unsupervised method that exploits truth's logical structure—a statement and its negation cannot both be true—to recover latent… Burns et al. · arXiv · 2022

LLMs & generative AIFact-checking & verification Atlas: Few-shot Learning with Retrieval Augmented Language ModelsATLAS is a retrieval-augmented language model that achieves strong few-shot performance on knowledge-intensive tasks by jointly training a dense retriever with a sequence-to-sequence language model. Unlike standard language models that must memorize knowledge, ATLAS retrieves relevant documents at both pre-training and fine-tuning… Izacard et al. · arXiv · 2022

Platforms, policy & societyDetection methods A Comprehensive Survey on Trustworthy Recommender SystemsRecommender systems, while beneficial to users, can amplify harmful content, introduce discrimination, and enable adversarial attacks. This survey provides a comprehensive framework for trustworthy recommender systems across six critical dimensions: Safety & Robustness (defense against adversarial attacks), Non-discrimination & Fairness… Fan et al. · ACM Computing Surveys · 2022

LLMs & generative AIDetection methods Causal Machine Learning: A Survey and Open ProblemsComprehensive survey of causal machine learning (CausalML), which formalizes data generation as a structural causal model to reason about interventions and counterfactuals. Categorizes 191 pages of CausalML work into five groups—causal supervised learning, causal generative modeling, causal explanations, causal fairness, and causal… Kaddour et al. · arXiv · 2022

LLMs & generative AIDatasets & evaluation Discovering Language Model Behaviors with Model-Written EvaluationsRather than manually creating evaluation datasets, this work proposes using language models themselves to generate high-quality evaluations for testing diverse LM behaviors. The authors generate 154 datasets testing behaviors spanning personality, political views, ethics, and goal-seeking. Perez et al. · 2022

LLMs & generative AIBots & coordinated campaigns Red Teaming Language Models with Language ModelsThis paper demonstrates that language models can automatically discover a wide range of harmful behaviors in other language models through "red teaming"—using one LM to generate adversarial test cases that probe another LM's vulnerabilities. The authors explore multiple generation strategies (zero-shot, few-shot, supervised learning,… Perez et al. · arXiv · 2022

Bots & coordinated campaignsDatasets & evaluation Botometer 101: Social bot practicum for computational social scientistsThis tutorial introduces Botometer, a supervised machine learning classifier for detecting social bots on Twitter. The authors explain how Botometer extracts features from account profiles and tweets, describe multiple versions optimized for different use cases (including BotometerLite for large-scale analysis), and provide practical… Yang et al. · arXiv · 2022

LLMs & generative AIDetection methods Real or Fake Text?: Investigating Human Ability to Detect Boundaries Between Human-Written and Machine-Generated TextDugan et al. investigate human ability to detect boundaries where text transitions from human-written to machine-generated, a more realistic and challenging task than binary fake/real classification. They introduce the RoFT (Real or Fake Text) game platform for crowdsourced annotation and release a dataset of 21,000+ annotations across… Dugan et al. · arXiv · 2022

Datasets & evaluationDetection methods NELA-GT-2022: A Large Multi-Labelled News Dataset for The Study of Misinformation in News ArticlesPresents the fifth NELA-GT dataset release: 1.78 million news articles from 361 outlets spanning all of 2022, with source-level veracity labels from Media Bias/Fact Check and embedded tweet data. Focuses on stabilizing collection infrastructure to achieve near-complete temporal coverage, enabling robust benchmarking of fake news… Gruppi et al. · arXiv · 2022