Tootfinder

ASTRA: Autonomous Spatial-Temporal Red-teaming for AI Software Assistants
Xiangzhe Xu, Guangyu Shen, Zian Su, Siyuan Cheng, Hanxi Guo, Lu Yan, Xuan Chen, Jiasheng Jiang, Xiaolong Jin, Chengpeng Wang, Zhuo Zhang, Xiangyu Zhang
https://arxiv.org/abs/2508.03936

ASTRA: Autonomous Spatial-Temporal Red-teaming for AI Software Assistants
AI coding assistants like GitHub Copilot are rapidly transforming software development, but their safety remains deeply uncertain-especially in high-stakes domains like cybersecurity. Current red-teaming tools often rely on fixed benchmarks or unrealistic prompts, missing many real-world vulnerabilities. We present ASTRA, an automated agent system designed to systematically uncover safety flaws in AI-driven code generation and security guidance systems. ASTRA works in three stages: (1) it build…

Peter Thiel's Armageddon
speaking tour has—like the world—not ended yet.
For a full two years now, the billionaire has been on the circuit,
spreading his biblically inflected ideas about doomsday
through a set of variably and sometimes visibly perplexed interviewers.
He has chatted onstage with the economist podcaster Tyler Cowen about the katechon
(the scriptural term for “that which withholds” the end times);
traded some very awkward on-camera…

The Real Stakes, and Real Story, of Peter Thiel’s Antichrist Obsession
Thirty years ago, a peace-loving Austrian theologian spoke to Peter Thiel about the apocalyptic theories of Nazi jurist Carl Schmitt. They’ve been a road map for the billionaire ever since.

When Hallucination Costs Millions: Benchmarking AI Agents in High-Stakes Adversarial Financial Markets
Zeshi Dai, Zimo Peng, Zerui Cheng, Ryan Yihe Li
https://arxiv.org/abs/2510.00332

When Hallucination Costs Millions: Benchmarking AI Agents in High-Stakes Adversarial Financial Markets
We present CAIA, a benchmark exposing a critical blind spot in AI evaluation: the inability of state-of-the-art models to operate in adversarial, high-stakes environments where misinformation is weaponized and errors are irreversible. While existing benchmarks measure task completion in controlled settings, real-world deployment demands resilience against active deception. Using crypto markets as a testbed where $30 billion was lost to exploits in 2024, we evaluate 17 models on 178 time-anchore…

Responding to the growing demand from organisations worldwide to reduce their dependency on US cloud providers, German software giant SAP has announced plans to invest more than €20 billion over the next decade in sovereign cloud infrastructure, research and development and personnel.

SAP promises €20bn+ to expand its sovereign cloud services
Responding to the growing demand from organisations worldwide to reduce their dependency on US cloud providers, German software giant SAP has announced plans to invest ...

A Primer on Causal and Statistical Dataset Biases for Fair and Robust Image Analysis
Charles Jones, Ben Glocker
https://arxiv.org/abs/2509.04295 https://ar…

A Primer on Causal and Statistical Dataset Biases for Fair and Robust Image Analysis
Machine learning methods often fail when deployed in the real world. Worse still, they fail in high-stakes situations and across socially sensitive lines. These issues have a chilling effect on the adoption of machine learning methods in settings such as medical diagnosis, where they are arguably best-placed to provide benefits if safely deployed. In this primer, we introduce the causal and statistical structures which induce failure in machine learning methods for image analysis. We highlight …

Ouste! (Yes it's a French word)
"French prime minister to face potential ousting in high-stakes confidence vote"
https://www.theguardian.com/world/2025/aug/25/francois-bayrou-france-prime-minister-confidence-vote

StockBench: Can LLM Agents Trade Stocks Profitably In Real-world Markets?
Yanxu Chen, Zijun Yao, Yantao Liu, Jin Ye, Jianing Yu, Lei Hou, Juanzi Li
https://arxiv.org/abs/2510.02209

StockBench: Can LLM Agents Trade Stocks Profitably In Real-world Markets?
Large language models (LLMs) have recently demonstrated strong capabilities as autonomous agents, showing promise in reasoning, tool use, and sequential decision-making. While prior benchmarks have evaluated LLM agents in domains such as software engineering and scientific discovery, the finance domain remains underexplored, despite its direct relevance to economic value and high-stakes decision-making. Existing financial benchmarks primarily test static knowledge through question answering, bu…

Adaptive Monitoring and Real-World Evaluation of Agentic AI Systems
Manish Shukla
https://arxiv.org/abs/2509.00115 https://arxiv.org/pdf/2509.00115

Adaptive Monitoring and Real-World Evaluation of Agentic AI Systems
Agentic artificial intelligence (AI) -- multi-agent systems that combine large language models with external tools and autonomous planning -- are rapidly transitioning from research laboratories into high-stakes domains. Our earlier "Basic" paper introduced a five-axis framework and proposed preliminary metrics such as goal drift and harm reduction but did not provide an algorithmic instantiation or empirical evidence. This "Advanced" sequel fills that gap. First, we revisit recent benchmarks a…

As a survivor of the type of stochastic terrorism that people like Charlie Kirk carry out, and also a survivor of gun violence, I've had a bit to think about.
I still understand the anger, the frustration, that leads someone to snap. I'll always have compassion, if also paired with some frustration, for the person who shot me and her husband. That makes sense to me. I refused to testify because it never made sense to punish a tool who acted out of ignorance.
What doesn't make as much sense to me are the upper class grifters, the stochastic terrorists, who turn a profit of off setting people like that up to kill. I always wanted to see people like that, people who were clearly profiting off of bringing evil into the world, held accountable.
But the way I always want to see them held accountable, is by being forced to live in a world where everyone is free. Charlie Kirk got out easy, and that is a bit sad. He should have had to suffer through our victory. He should have had to watch the fall of Western Civilization, the collapse of all the things he held dear, all the things he tried to uphold. I wish that he had lived long enough to understand the suffering he inflicted on the world. He should have lived long enough to have to do real work, to have to figure out how to feed himself, house himself, in a world that has no market for the hate that he brings.
As much as I had rather that he starved, I would never shed a tear for the opening of a new public urinal.
Rest in piss. Charlie Kirk.

Reinforcement Learning with Action-Triggered Observations
Alexander Ryabchenko, Wenlong Mou
https://arxiv.org/abs/2510.02149 https://arxiv.org/pdf/2510.021…

Reinforcement Learning with Action-Triggered Observations
We study reinforcement learning problems where state observations are stochastically triggered by actions, a constraint common in many real-world applications. This framework is formulated as Action-Triggered Sporadically Traceable Markov Decision Processes (ATST-MDPs), where each action has a specified probability of triggering a state observation. We derive tailored Bellman optimality equations for this framework and introduce the action-sequence learning paradigm in which agents commit to ex…