Tootfinder

Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets
Lei Hsiung, Tianyu Pang, Yung-Chen Tang, Linyue Song, Tsung-Yi Ho, Pin-Yu Chen, Yaoqing Yang
https://arxiv.org/abs/2506.05346

Why LLM Safety Guardrails Collapse After Fine-tuning: A Similarity Analysis Between Alignment and Fine-tuning Datasets
Recent advancements in large language models (LLMs) have underscored their vulnerability to safety alignment jailbreaks, particularly when subjected to downstream fine-tuning. However, existing mitigation strategies primarily focus on reactively addressing jailbreak incidents after safety guardrails have been compromised, removing harmful gradients during fine-tuning, or continuously reinforcing safety alignment throughout fine-tuning. As such, they tend to overlook a critical upstream factor: …

Buyer With Ties to Chinese Communist Party Got V.I.P. Treatment at Trump Crypto Dinner (New York Times)
https://www.nytimes.com/2025/06/06/us/politics/trump-crypto-dinner-china-he-tianying.html
http://www.memeorandum.com/250606/p144#a250606p144

The Chinese Pulsar Timing Array data release I. Single pulsar noise analysis
Siyuan Chen, Heng Xu, Yanjun Guo, Bojun Wang, R. Nicolas Caballero, Jinchen Jiang, Jiangwei Xu, Zihan Xue, Kejia Lee, Jianping Yuan, Yonghua Xu, Jingbo Wang, Longfei Hao, Jintao Luo, Jinlin Han, Peng Jiang, Zhiqiang Shen, Min Wang, Na Wang, Renxin Xu, Xiangping Wu, Lei Qian, Xin Guan, Menglin Huang, Chun Sun, Yan Zhu

The Chinese Pulsar Timing Array data release I. Single pulsar noise analysis
The Chinese Pulsar Timing Array (CPTA) has collected observations from 57 millisecond pulsars using the Five-hundred-meter Aperture Spherical Radio Telescope (FAST) for close to three years, for the purpose of searching for gravitational waves (GWs). To robustly search for ultra-low-frequency GWs, pulsar timing arrays (PTAs) need to use models to describe the noise from the individual pulsars. We report on the results from the single pulsar noise analysis of the CPTA data release I (DR1). Conve…

Gradient Inversion Attacks on Parameter-Efficient Fine-Tuning
Hasin Us Sami, Swapneel Sen, Amit K. Roy-Chowdhury, Srikanth V. Krishnamurthy, Basak Guler
https://arxiv.org/abs/2506.04453

Gradient Inversion Attacks on Parameter-Efficient Fine-Tuning
Federated learning (FL) allows multiple data-owners to collaboratively train machine learning models by exchanging local gradients, while keeping their private data on-device. To simultaneously enhance privacy and training efficiency, recently parameter-efficient fine-tuning (PEFT) of large-scale pretrained models has gained substantial attention in FL. While keeping a pretrained (backbone) model frozen, each user fine-tunes only a few lightweight modules to be used in conjunction, to fit speci…

🇺🇦 #NowPlaying on KEXP's #MiddayShow
Tomo Nakayama:
🎵 Get To Know You
#TomoNakayama
https://tomomusic.bandcamp.com/track/get-to-know-you
https://open.spotify.com/track/6LpfuMXJuFc1WQbv9xqizX

CO-RFT: Efficient Fine-Tuning of Vision-Language-Action Models through Chunked Offline Reinforcement Learning
Dongchi Huang, Zhirui Fang, Tianle Zhang, Yihang Li, Lin Zhao, Chunhe Xia
https://arxiv.org/abs/2508.02219

CO-RFT: Efficient Fine-Tuning of Vision-Language-Action Models through Chunked Offline Reinforcement Learning
Vision-Language-Action (VLA) models demonstrate significant potential for developing generalized policies in real-world robotic control. This progress inspires researchers to explore fine-tuning these models with Reinforcement Learning (RL). However, fine-tuning VLA models with RL still faces challenges related to sample efficiency, compatibility with action chunking, and training stability. To address these challenges, we explore the fine-tuning of VLA models through offline reinforcement lear…

Precise Timing Analysis of Four Magnetic Cataclysmic Variables with TESS
Srinivas M Rao, Jeewan C Pandey, Nikita Rawat, Arti Joshi, Ajay Kumar Singh
https://arxiv.org/abs/2506.04371

Precise Timing Analysis of Four Magnetic Cataclysmic Variables with TESS
We analysed high time-resolution optical photometric data from the Transiting Exoplanet Survey Satellite (TESS) to study the timing behaviour of four intermediate polar-like objects, namely, V1460 Her, 1RXS J045707.4+452751, Swift J0958.0-4208, and V842 Cen. In the case of V1460 Her, we refined the measurement of its orbital period. Long-term observations suggest a gradual decrease in the orbital period of V1460 Her, and the stable light curve during the TESS observations indicates its quiescen…

This is a good analogy for AI, if you don't like toast and that's not jelly.

ROSGuard: A Bandwidth Regulation Mechanism for ROS2-based Applications
Jon Altonaga Puente, Enrico Mezzetti, Irune Agirre Troncoso, Jaume Abella Ferrer, Francisco J. Cazorla Almeida
https://arxiv.org/abs/2506.04640

ROSGuard: A Bandwidth Regulation Mechanism for ROS2-based Applications
Multicore timing interference, arising when multiple requests contend for the same shared hardware resources, is a primary concern for timing verification and validation of time-critical applications. Bandwidth control and regulation approaches have been proposed in the literature as an effective method to monitor and limit the impact of timing interference at run time. These approaches seek for fine-grained control of the bandwidth consumption (at the microsecond level) to meet stringent timin…

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…