Tootfinder

This https://arxiv.org/abs/2502.13172 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csCR_…

Unveiling Privacy Risks in LLM Agent Memory
Large Language Model (LLM) agents have become increasingly prevalent across various real-world applications. They enhance decision-making by storing private user-agent interactions in the memory module for demonstrations, introducing new privacy risks for LLM agents. In this work, we systematically investigate the vulnerability of LLM agents to our proposed Memory EXTRaction Attack (MEXTRA) under a black-box setting. To extract private information from memory, we propose an effective attacking …

This https://arxiv.org/abs/2505.23419 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csSE_…

SWE-bench Goes Live!
The issue-resolving task, where a model generates patches to fix real-world bugs, has emerged as a critical benchmark for evaluating the capabilities of large language models (LLMs). While SWE-bench and its variants have become standard in this domain, they suffer from key limitations: they have not been updated since their initial releases, cover a narrow set of repositories, and depend heavily on manual effort for instance construction and environment setup. These factors hinder scalability a…

Augment or Not? A Comparative Study of Pure and Augmented Large Language Model Recommenders
Wei-Hsiang Huang, Chen-Wei Ke, Wei-Ning Chiu, Yu-Xuan Su, Chun-Chun Yang, Chieh-Yuan Cheng, Yun-Nung Chen, Pu-Jen Cheng
https://arxiv.org/abs/2505.23053

Augment or Not? A Comparative Study of Pure and Augmented Large Language Model Recommenders
Large language models (LLMs) have introduced new paradigms for recommender systems by enabling richer semantic understanding and incorporating implicit world knowledge. In this study, we propose a systematic taxonomy that classifies existing approaches into two categories: (1) Pure LLM Recommenders, which rely solely on LLMs, and (2) Augmented LLM Recommenders, which integrate additional non-LLM techniques to enhance performance. This taxonomy provides a novel lens through which to examine the …

This https://arxiv.org/abs/2505.21938 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csLG_…

Practical Adversarial Attacks on Stochastic Bandits via Fake Data Injection
Adversarial attacks on stochastic bandits have traditionally relied on some unrealistic assumptions, such as per-round reward manipulation and unbounded perturbations, limiting their relevance to real-world systems. We propose a more practical threat model, Fake Data Injection, which reflects realistic adversarial constraints: the attacker can inject only a limited number of bounded fake feedback samples into the learner's history, simulating legitimate interactions. We design efficient attack …

Turns out that if you model online spaces after real world ones, it works pretty well. Online spaces let you be in many places at once which changes the dynamics, but as an example of something that ports well,
https://www.patternlanguageindex.com/patterns/intimacy-gradient
This pattern works not just for the design of houses but online spaces. Let people get to know a group in less-intimate space before they end up in the more-intimate space. Having a few gradations works really well, and it doesn't have to be a power play or status game.

Impact of Tariff Wars on Global Economy
N. S. Gonchar, O. P. Dovzhyk, A. S. Zhokhin, W. H. Kozyrsky, A. P. Makhort
https://arxiv.org/abs/2505.05576 https:/…

Impact of Tariff Wars on Global Economy
The Ricardian model of world trade based on comparative advantage is not sufficient to justify equal trade relations.The existing model of trade relations does not explain the distribution of income among trading countries. This paper presents a method for building equitable trade relations. Its essence is to present an algorithm for building such trade relations, based on the previously proposed model of world trade, that the trade balance of each country would be equal to zero. Under such con…

RoboEgo System Card: An Omnimodal Model with Native Full Duplexity
Yiqun Yao, Xiang Li, Xin Jiang, Xuezhi Fang, Naitong Yu, Aixin Sun, Yequan Wang
https://arxiv.org/abs/2506.01934

RoboEgo System Card: An Omnimodal Model with Native Full Duplexity
Humans naturally process real-world multimodal information in a full-duplex manner. In artificial intelligence, replicating this capability is essential for advancing model development and deployment, particularly in embodied contexts. The development of multimodal models faces two primary challenges: (1) effectively handling more than three modalities-such as vision, audio, and text; and (2) delivering full-duplex responses to rapidly evolving human instructions. To facilitate research on mode…

This https://arxiv.org/abs/2505.08944 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csDC_…

Toward Cost-Efficient Serving of Mixture-of-Experts with Asynchrony
Mixture-of-Experts (MoE) architectures offer the promise of larger model capacity without the prohibitive costs of fully dense designs. However, in real-world inference serving, load skew across experts often leads to suboptimal device utilization and excessive synchronization overheads. This paper introduces Asynchronous Expert Parallelism (AEP), a new paradigm that decouples layer execution from barrier-style synchronization. By dynamically queuing tokens at each layer (referred to as $μ$-qu…

This https://arxiv.org/abs/2505.21906 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csRO_…

Vision-Language-Action Model with Open-World Embodied Reasoning from Pretrained Knowledge
Vision-language-action (VLA) models have emerged as the next generation of models in robotics. However, despite leveraging powerful pre-trained Vision-Language Models (VLMs), existing end-to-end VLA systems often lose key capabilities during fine-tuning as the model adapts to specific robotic tasks. We argue that a generalizable VLA model should retain and expand upon the VLM's core competencies: 1) Open-world embodied reasoning - the VLA should inherit the knowledge from VLM, i.e., recognize a…

I read "Then I Am Myself the World: What Consciousness Is and How to Expand It" by Christof Koch.
Interesting book which spends like 8 or 9 chapters detailing all the experiments which prove beyond much doubt that consciousness, and self awareness, is a thing done by a brain.
It describes how perception is a construction of a description, has a chapter called "computational mind"
And then spends the last two chapters describing why he thinks the mind can't be computed, because drugs have made him think experience is some kind of magic associated with highly interconnected causal structures.
Apparently, he thinks, once things become interconnected enough they become able to cause things independently of the physics running those connections.
Which is crazy, obviously. There's nothing causal in direct connections between neurons that isn't equally causal in modeled connections between virtual neurons.
All his evidence in the book from neural MRI scans to the effects of psychedelic drugs and symptoms of strokes and disease point to the brain simulating a virtual reality which is the basis of perception.
That simulated world in which we live is full of colour and shape and sounds and emotions and millions of mental constructs that are built to be correlated by the senses with the outside world, but are not equal to the world itself. We live in a dream constructed to correlate with reality.
But then instead of taking the next step: That consciousness itself is a property of a simulated being inside that mental model of the universe, a property which the brain simulates and applies to the virtual self that's doing the experiencing inside that model, he jumps towards some magic implying pan-psychism or that sufficiently interconnected networks become causally self-complete for some reason nobody can fathom.
Sure, colour and shape and emotions are all made up by the brain but experience can't be! For some reason.
You see in truth dualism is false, in that there is no spirit realm in which ghosts animate the matter of the body somehow.
Yet also, dualism is true, in that there is a simulated mental reality which we live in, computed by the brain in which all perception and experience are created, which is related-to but separate-from the unfolding complicated dance of energy that is the universe our bodies interact with.
People take some DMT trip, and the model of the universe emulated by their brain collapses and breaks. Their virtual simulated self inside their mind has these experiences of being one with the universe or the experience of feeling dead yet conscious or whatever, and these hippies think that the broken down simulated experience is real and reflects how consciousness is more fundamental than the atoms that make up the neurons in their brain.
Instead of realizing it shows them that their experienced universe is a simulacrum, they think they get a more direct experience of reality somehow. A consciousness more pure than any mere base atom.
"Then I am myself the world" is a great title. Everything you ever experience is created and simulated in your brain like a dream, the whole universe is inside your head. Even the fact of experience itself.
But that isn't the conclusion Koch reaches somehow, he just jumps from describing the evidence that this is so straight into ascribing super-causal magic consciousness to particular arrangements of atoms that integrated information theory suggest have high correlation, and thinks therefore conciousness is itself the entire universe.
Ah well, fun book. I like arguing in my head with authors that are wrong.
#reading #books #consciousness #thenIAmMyselfTheWorld

Conformance Checking for Less: Efficient Conformance Checking for Long Event Sequences
Eli Bogdanov, Izack Cohen, Avigdor Gal
https://arxiv.org/abs/2505.21506

Conformance Checking for Less: Efficient Conformance Checking for Long Event Sequences
Long event sequences (termed traces) and large data logs that originate from sensors and prediction models are becoming increasingly common in our data-rich world. In such scenarios, conformance checking-validating a data log against an expected system behavior (the process model) can become computationally infeasible due to the exponential complexity of finding an optimal alignment. To alleviate scalability challenges for this task, we propose ConLES, a sliding-window conformance checking appr…

This https://arxiv.org/abs/2505.21432 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csRO_…

Hume: Introducing System-2 Thinking in Visual-Language-Action Model
Humans practice slow thinking before performing actual actions when handling complex tasks in the physical world. This thinking paradigm, recently, has achieved remarkable advancement in boosting Large Language Models (LLMs) to solve complex tasks in digital domains. However, the potential of slow thinking remains largely unexplored for robotic foundation models interacting with the physical world. In this work, we propose Hume: a dual-system Vision-Language-Action (VLA) model with value-guided…

This https://arxiv.org/abs/2502.03663 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csDS_…

Fast Geographic Routing in Fixed-Growth Graphs
In the 1960s, the social scientist Stanley Milgram performed his famous "small-world" experiments where he found that people in the US who are far apart geographically are nevertheless connected by remarkably short chains of acquaintances. Since then, there has been considerable work to design networks that accurately model the phenomenon that Milgram observed. One well-known approach was Barab{á}si and Albert's preferential attachment model, which has small diameter yet lacks an algorithm tha…

This https://arxiv.org/abs/2503.03480 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csRO_…

SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning
Vision-language-action models (VLAs) show potential as generalist robot policies. However, these models pose extreme safety challenges during real-world deployment, including the risk of harm to the environment, the robot itself, and humans. How can safety constraints be explicitly integrated into VLAs? We address this by exploring an integrated safety approach (ISA), systematically modeling safety requirements, then actively eliciting diverse unsafe behaviors, effectively constraining VLA poli…

SWE-bench Goes Live!
Linghao Zhang, Shilin He, Chaoyun Zhang, Yu Kang, Bowen Li, Chengxing Xie, Junhao Wang, Maoquan Wang, Yufan Huang, Shengyu Fu, Elsie Nallipogu, Qingwei Lin, Yingnong Dang, Saravan Rajmohan, Dongmei Zhang
https://arxiv.org/abs/2505.23419

SWE-bench Goes Live!
The issue-resolving task, where a model generates patches to fix real-world bugs, has emerged as a critical benchmark for evaluating the capabilities of large language models (LLMs). While SWE-bench and its variants have become standard in this domain, they suffer from key limitations: they have not been updated since their initial releases, cover a narrow set of repositories, and depend heavily on manual effort for instance construction and environment setup. These factors hinder scalability a…

Clustering scientific publications: lessons learned through experiments with a real citation network
Vu Thi Huong, Thorsten Koch
https://arxiv.org/abs/2505.18180

Clustering scientific publications: lessons learned through experiments with a real citation network
Clustering scientific publications can reveal underlying research structures within bibliographic databases. Graph-based clustering methods, such as spectral, Louvain, and Leiden algorithms, are frequently utilized due to their capacity to effectively model citation networks. However, their performance may degrade when applied to real-world data. This study evaluates the performance of these clustering algorithms on a citation graph comprising approx. 700,000 papers and 4.6 million citations ex…

This https://arxiv.org/abs/2505.12583 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csRO_…

A Comprehensive Survey on Physical Risk Control in the Era of Foundation Model-enabled Robotics
Recent Foundation Model-enabled robotics (FMRs) display greatly improved general-purpose skills, enabling more adaptable automation than conventional robotics. Their ability to handle diverse tasks thus creates new opportunities to replace human labor. However, unlike general foundation models, FMRs interact with the physical world, where their actions directly affect the safety of humans and surrounding objects, requiring careful deployment and control. Based on this proposition, our survey co…

This https://arxiv.org/abs/2501.07985 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csRO_…

CHEQ-ing the Box: Safe Variable Impedance Learning for Robotic Polishing
Robotic systems are increasingly employed for industrial automation, with contact-rich tasks like polishing requiring dexterity and compliant behaviour. These tasks are difficult to model, making classical control challenging. Deep reinforcement learning (RL) offers a promising solution by enabling the learning of models and control policies directly from data. However, its application to real-world problems is limited by data inefficiency and unsafe exploration. Adaptive hybrid RL methods blen…