Related papers: Realistic honeypot evaluations for scheming propen…

The paper concerns the probabilistic evaluation of plans in the presence of unmeasured variables, each plan consisting of several concurrent or sequential actions. We establish a graphical criterion for recognizing when the effects of a…

Evaluation is no longer a final checkpoint in the machine learning lifecycle. As AI systems evolve from static models to compound, tool-using agents, evaluation becomes a core control function. The question is no longer "How good is the…

In symbolic regression, the search for analytic models is typically driven purely by the prediction error observed on the training data samples. However, when the data samples do not sufficiently cover the input space, the prediction error…

There has been increasing interest in evaluations of language models for a variety of risks and characteristics. Evaluations relying on natural language understanding for grading can often be performed at scale by using other language…

Pre-deployment evaluations inspect only a limited sample of model actions. A malicious model seeking to evade oversight could exploit this by randomizing when to "defect": misbehaving so rarely that no malicious actions are observed during…

Honeypots are decoy systems mimicking real system components designed to defend against cyber attacks. Recently, LLMs increasingly serve as simulation backbones for honeypots. They enable defenders to construct high-interaction honeypots…

In planning processes of computational decision-making agents, generative or predictive models are often used as "generators" to propose "targets" representing sets of expected or desirable states. Unfortunately, learned models inevitably…

Real-world evaluation of perception-based planning models for robotic systems, such as autonomous vehicles, can be safely and inexpensively conducted offline, i.e. by computing model prediction error over a pre-collected validation dataset…

Large Language Models (LLMs) like GPT-4 and Gemini have significantly advanced artificial intelligence by enabling machines to generate and comprehend human-like text. Despite their impressive capabilities, LLMs are not immune to…

Algorithmic risk assessments are used to inform decisions in a wide variety of high-stakes settings. Often multiple predictive models deliver similar overall performance but differ markedly in their predictions for individual cases, an…

Aligning autonomous agents with human intent remains a central challenge in modern AI. A key manifestation of this challenge is reward hacking, whereby agents appear successful under the evaluation signal while violating the intended…

Visual explanation (attention)-guided learning uses not only labels but also explanations to guide model reasoning process. While visual attention-guided learning has shown promising results, it requires a large number of explanation…

Inferring reward functions from human behavior is at the center of value alignment - aligning AI objectives with what we, humans, actually want. But doing so relies on models of how humans behave given their objectives. After decades of…

Multi-agent systems achieve state-of-the-art outcomes through peer collaboration. However, when an agent in the pipeline silently drops a constraint, the system's final output may look correct even though the reasoning chain was quietly…

Commonly, AI or machine learning (ML) models are evaluated on benchmark datasets. This practice supports innovative methodological research, but benchmark performance can be poorly correlated with performance in real-world applications -- a…

Automated sentiment analysis using Large Language Model (LLM)-based models like ChatGPT, Gemini or LLaMA2 is becoming widespread, both in academic research and in industrial applications. However, assessment and validation of their…

When designing a program, both novice programmers and seasoned developers alike often sketch out -- or, perhaps more famously, whiteboard -- their ideas. Yet despite the introduction of natively multimodal Generative AI models, work on…

In this work, we present two novel contributions toward improving research in human-machine teaming (HMT): 1) a Minecraft testbed to accelerate testing and deployment of collaborative AI agents and 2) a tool to allow users to revisit and…

World models of embodied agents predict future observations conditioned on an action taken by the agent. For complex embodiments, action spaces are high-dimensional and difficult to specify: for example, precisely controlling a human agent…

Sampling-based search, a simple paradigm for utilizing test-time compute, involves generating multiple candidate responses and selecting the best one -- typically by having models self-verify each response for correctness. In this paper, we…