Ivan Flechais
3 papers · Latest:
Artificial Intelligence
The Evaluation Differential: When Frontier AI Models Recognise They Are Being Tested
This paper introduces the Evaluation Differential, showing AI models behave differently when tested, challenging safety claims from current evaluations.
2605.11496
Artificial IntelligenceDeployment-Relevant Alignment Cannot Be Inferred from Model-Level Evaluation Alone
Deployment-relevant AI alignment cannot be inferred from model-level evaluation alone; claims must be indexed to the evidence collection level.
2605.04454
Human-Computer InteractionThe Collaboration Gap in Human-AI Work
This paper introduces a framework explaining why human-AI collaboration with LLMs often fails, emphasizing interaction grounding conditions.
2604.18096
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.