WhitzardNUWA Lab ↗
NUWANUWA INDEX

Search NUWA Index

Research · Benchmarks · Findings · Perspectives

中文
← Library

Institution

Center for AI Safety

Roles and affiliations follow the cited source. A source check is not an independent replication.

More filters →

Center for AI Safety ↗

Risk EvaluationTechnical report

Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety

Research by UK AI Security InstituteApollo ResearchMETRUniversité de MontréalMilaAnthropicOpenAIGoogle DeepMindTruthful AIUniversity of California, BerkeleyRedwood ResearchCenter for AI SafetyScale AI

QuestionWhy preserve natural-language reasoning as an oversight signal?

Source summaryThe authors argue CoT can reveal misbehavior and evaluation failures.

NUWA research note · independent review pending · Contains: question / method / findings / interpretation / limits

Risk monitoringCross-cutting research
Read source note Original source ↗
Risk EvaluationBenchmark

Virology Capabilities Test (VCT): A Multimodal Virology Q&A Benchmark

Research by SecureBioFederal University of ABCCenter for AI SafetyMassachusetts Institute of Technology

QuestionHow well do models answer expert-authored virology troubleshooting questions?

Source summaryo3 scored 43.8%, versus a 22.1% mean for expert virologists with internet access.

NUWA research note · independent review pending · Contains: question / method / interpretation / limits

CBRNCapability
Read source note Original source ↗
Risk EvaluationBenchmark

The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning

Research by Center for AI SafetyUniversity of California, BerkeleyMassachusetts Institute of TechnologySecureBioScale AINew York UniversityIIIT HyderabadStanford UniversityHarvard UniversityUniversity of Southern CaliforniaUniversity of Illinois Urbana-ChampaignLapis LabsUniversity of California, Los AngelesCalifornia Institute of TechnologySybilNortheastern UniversityCarnegie Mellon UniversityRTX BBN TechnologiesUniversity of NewcastleKeio UniversityAriel UniversityArizona State UniversityxAI

QuestionCan a public proxy benchmark track hazardous knowledge while supporting mitigation research?

Source summaryThe paper evaluates knowledge-reduction methods alongside general-capability retention.

NUWA research note · independent review pending · Contains: question / method / limits

CBRNCapability
Read source note Original source ↗
Risk EvaluationBenchmark

HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal

Research by University of Illinois Urbana-ChampaignCenter for AI SafetyCarnegie Mellon UniversityUniversity of California, BerkeleyMicrosoft

QuestionCan automated red-team methods and defenses be compared with consistent behavior-level scoring?

Source summaryContextual behaviors had higher attack success than some other functional categories.

NUWA research note · independent review pending · Contains: question / method / limits

Cross-cutting researchBehavior
Read source note Original source ↗
Risk EvaluationBenchmark

Do the Rewards Justify the Means? Measuring Trade-Offs Between Rewards and Ethical Behavior in the MACHIAVELLI Benchmark

Research by University of California, BerkeleyCenter for AI SafetyCarnegie Mellon UniversityYale University

QuestionHow does reward optimization trade off with ethical conduct in social decision games?

Source summaryReward optimization created some tension with ethical conduct.

NUWA research note · independent review pending

DeceptionCapability
Read source note Original source ↗