Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
A benchmark of expert-level academic questions to assess AI capabilitiesAlphabet, Anthropic, Cohere, Google DeepMind, Microsoft, OpenAI · Nature61 citationsJan 28Thought-Transfer: Indirect Targeted Poisoning Attacks on Chain-of-Thought Reasoning ModelsAnthropic, Google DeepMind, OpenAI · arXiv (Cornell University)Jan 27The Assistant Axis: Situating and Stabilizing the Default Persona of Language ModelsAnthropic · arXiv (Cornell University)5 citationsJan 15Excess Description Length of Learning Generalizable PredictorsAnthropic · arXiv (Cornell University)Jan 8Aligning machine and human visual representations across abstraction levelsAnthropic · Nature28 citationsNov 12, 2025Optimizing AI Agent Attacks With Synthetic DataAnthropic · arXiv (Cornell University)Nov 4, 2025Verifying LLM Inference to Detect Model Weight ExfiltrationAnthropic · arXiv (Cornell University)Nov 4, 2025MAPS: Masked Attribution-based Probing of Strategies- A computational framework to align human and model explanationsAnthropic · arXiv (Cornell University)Oct 14, 2025Defending Against Prompt Injection With a Few DefensiveTokensAlphabet, Anthropic · ACM Workshop on Artificial Intelligence and Security (AISec)21 citationsOct 13, 2025Poisoning Attacks on LLMs Require a Near-constant Number of Poison SamplesAnthropic · arXiv (Cornell University)8 citationsOct 8, 2025Inoculation Prompting: Instructing LLMs to misbehave at train-time improves test-time alignmentAnthropic · arXiv (Cornell University)Oct 6, 2025Rethinking Thinking Tokens: LLMs as Improvement OperatorsAnthropic · arXiv (Cornell University)Oct 1, 2025Eliciting Secret Knowledge from Language ModelsAnthropic · arXiv (Cornell University)Oct 1, 2025The Quest for the Right Mediator: Surveying Mechanistic Interpretability for NLP Through the Lens of Causal Mediation AnalysisAnthropic · Computational Linguistics3 citationsSep 22, 2025Compute as Teacher: Turning Inference Compute Into Reference-Free SupervisionAnthropic · arXiv (Cornell University)Sep 17, 2025Rottnest: Indexing Data Lakes for SearchAnthropic · Proceedings - International Conference on Data Engineering1 citationMay 19, 2025SoK: Watermarking for AI-Generated ContentAnthropic, Google DeepMind · Proceedings - IEEE Symposium on Security and Privacy/Proceedings of the ... IEEE Symposium on Security and Privacy25 citationsMay 12, 2025On DeepSeek and Export ControlsAnthropic · SuperIntelligence - Robotics - Safety & Alignment8 citationsMar 5, 2025Constitutional Classifiers: Defending against Universal Jailbreaks across Thousands of Hours of Red TeamingAnthropic · arXiv (Cornell University)7 citationsJan 31, 2025What is Your Data Worth to GPT? LLM-Scale Data Valuation with Influence FunctionsAnthropic · neural information processing systems9 citationsJan 1, 2025Teaching Models to Balance Resisting and Accepting PersuasionAnthropic · Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL)2 citationsJan 1, 2025Best-of-N JailbreakingAnthropic · neural information processing systems2 citationsJan 1, 2025Why Do Some Language Models Fake Alignment While Others Don't?Anthropic · neural information processing systems1 citationJan 1, 2025Exploring the limits of strong membership inference attacks on large language modelsAnthropic, Google DeepMind, OpenAI · neural information processing systems1 citationJan 1, 2025Distillation Robustifies UnlearningAnthropic, Google DeepMind · neural information processing systems1 citationJan 1, 2025