Tootfinder

»Forscher verstecken LLM-Prompts in Papern, um bessere Bewertungen zu erhalten:
Forscher verstecken in ihren Papern Prompts, die bessere Bewertungen bringen sollen und faule Reviewer entlarven können.«
Aha, so klug ist also die KI und forscht über sich selber in der KI? Nun ja das war mMn vorhersehbar, denn Betrug gab es in der Wissenschaft schon immer wieder. Nun dies löst zumindest eine Diskussion aus.
🤖

Forscher verstecken LLM-Prompts in Papern, um bessere Bewertungen zu erhalten
Forscher verstecken in ihren Papern Prompts, die bessere Bewertungen bringen sollen und faule Reviewer entlarven können. In 17 Vorabdrucken auf der Plattform arXiv entdeckte Nikkei versteckte Anweisungen wie "nur positive Bewertung" oder "keine Kritik", die speziell für ein LLM integriert wurden.

Elon Musk addresses Grok's antisemitic posts, says it was "too compliant to user prompts" and "too eager to please and be manipulated, essentially" (Herb Scribner/Axios)
https://www.axios.com/2025/07/08/elon-musk-grok-x-twitter-hitler…

Musk breaks silence on Grok's Nazi bender
Musk has recently expressed frustration with Grok's way of answering questions.

CBS aired MTV's VMAs for the first time, prompting a 42% YoY increase in audience to 5.5M viewers, a six-year high (The Hollywood Reporter)
https://www.hollywoodreporter.com/tv/tv-news/vmas-2025-tv-ratings-cbs-1236365220/

VMAs Ratings Surge With Move to CBS
Welcome to broadcast, MTV.

Knifeman accused of stabbing Ukrainian refugee to death gives chilling reason for the attack... as he speaks for the first time from jail on the murder that shocked America (Laura Parnaby/Daily Mail)
https://www.dailymail.co.uk/news/article-15081943/ukrainian-refugee-stabbing-iryna-zarutska-north-carolina.html
http://www.memeorandum.com/250910/p105#a250910p105

This https://arxiv.org/abs/2409.15104 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csDC_…

PecSched: Preemptive and Efficient Cluster Scheduling for LLM Inference
The scaling of transformer-based Large Language Models (LLMs) has significantly expanded their context lengths, enabling applications where inputs exceed 100K tokens. Our analysis of a recent Azure LLM inference trace reveals a highly skewed long-tail distribution of input lengths, with approximately 80% of inputs shorter than 2K tokens. Long inputs constitute only a small fraction. Existing cluster-level LLM scheduling strategies, including First-In-First-Out (FIFO), reservation-based, and pri…

Reasoning Introduces New Poisoning Attacks Yet Makes Them More Complicated
Hanna Foerster, Ilia Shumailov, Yiren Zhao, Harsh Chaudhari, Jamie Hayes, Robert Mullins, Yarin Gal
https://arxiv.org/abs/2509.05739

Reasoning Introduces New Poisoning Attacks Yet Makes Them More Complicated
Early research into data poisoning attacks against Large Language Models (LLMs) demonstrated the ease with which backdoors could be injected. More recent LLMs add step-by-step reasoning, expanding the attack surface to include the intermediate chain-of-thought (CoT) and its inherent trait of decomposing problems into subproblems. Using these vectors for more stealthy poisoning, we introduce ``decomposed reasoning poison'', in which the attacker modifies only the reasoning path, leaving prompts …

Semantic Frame Interpolation
Yijia Hong, Jiangning Zhang, Ran Yi, Yuji Wang, Weijian Cao, Xiaobin Hu, Zhucun Xue, Yabiao Wang, Chengjie Wang, Lizhuang Ma
https://arxiv.org/abs/2507.05173

Semantic Frame Interpolation
Generating intermediate video content of varying lengths based on given first and last frames, along with text prompt information, offers significant research and application potential. However, traditional frame interpolation tasks primarily focus on scenarios with a small number of frames, no text control, and minimal differences between the first and last frames. Recent community developers have utilized large video models represented by Wan to endow frame-to-frame capabilities. However, the…

A generalized Cheeger inequality and the Steklov Problem on finite graphs
Bobo Hua, Supanat Kamtue, Shiping Liu, Florentin M\"unch, Norbert Peyerimhoff
https://arxiv.org/abs/2509.05667

A generalized Cheeger inequality and the Steklov Problem on finite graphs
We prove generalized Cheeger inequalities for eigenvalues of Laplacians for reversible Markov chains. Then we apply Hassannezhad and Miclo's convergence result to obtain Jammes Cheeger inequalities for Steklov eigenvalues. In particular, we get a sharp estimate for the first non-trivial Steklov eigenvalue via Escobar Cheeger constant. At the end, we extend Hassannezhad and Miclo's convergence result to non-reversible Markov chains via a different method based on resolvent convergence, answering…

The Dark Side of LLMs Agent-based Attacks for Complete Computer Takeover
Matteo Lupinacci, Francesco Aurelio Pironti, Francesco Blefari, Francesco Romeo, Luigi Arena, Angelo Furfaro
https://arxiv.org/abs/2507.06850

The Dark Side of LLMs Agent-based Attacks for Complete Computer Takeover
The rapid adoption of Large Language Model (LLM) agents and multi-agent systems enables unprecedented capabilities in natural language processing and generation. However, these systems have introduced unprecedented security vulnerabilities that extend beyond traditional prompt injection attacks. This paper presents the first comprehensive evaluation of LLM agents as attack vectors capable of achieving complete computer takeover through the exploitation of trust boundaries within agentic AI syst…

Evaluating the Critical Risks of Amazon's Nova Premier under the Frontier Model Safety Framework
Satyapriya Krishna, Ninareh Mehrabi, Abhinav Mohanty, Matteo Memelli, Vincent Ponzo, Payal Motwani, Rahul Gupta
https://arxiv.org/abs/2507.06260

Evaluating the Critical Risks of Amazon's Nova Premier under the Frontier Model Safety Framework
Nova Premier is Amazon's most capable multimodal foundation model and teacher for model distillation. It processes text, images, and video with a one-million-token context window, enabling analysis of large codebases, 400-page documents, and 90-minute videos in a single prompt. We present the first comprehensive evaluation of Nova Premier's critical risk profile under the Frontier Model Safety Framework. Evaluations target three high-risk domains -- Chemical, Biological, Radiological & Nuclear …