Tootfinder

This https://arxiv.org/abs/2506.02965 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csLG_…

Memory-Efficient and Privacy-Preserving Collaborative Training for Mixture-of-Experts LLMs
Mixture-of-Experts (MoE) has been gaining popularity due to its successful adaptation to large language models (LLMs). In this work, we introduce Privacy-preserving Collaborative Mixture-of-Experts (PC-MoE), which leverages the sparsity of the MoE architecture for memory-efficient decentralized collaborative LLM training, enabling multiple parties with limited GPU-memory and data resources to collectively train more capable LLMs than they could achieve individually. At the same time, this appro…

Myopic non-intersection in a periodic potential
Jonas Arista, Daniel Remenik, Avelio Sep\'ulveda
https://arxiv.org/abs/2506.05246 https://

Myopic non-intersection in a periodic potential
We introduce a class of Markov processes conditioned to avoid intersection over a moving time window of length T>0, a setting we refer to as myopic non-intersection. In particular, we study a system of myopic non-intersecting Brownian motions subject to a periodic potential. Our focus lies in understanding the interplay between the confining effect of the potential and the repulsion induced by the non-intersection constraint. We show that, in the long time limit, and as both T and the strength …

This https://arxiv.org/abs/2506.02040 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csCR_…

Beyond the Protocol: Unveiling Attack Vectors in the Model Context Protocol Ecosystem
The Model Context Protocol (MCP) is an emerging standard designed to enable seamless interaction between Large Language Model (LLM) applications and external tools or resources. Within a short period, thousands of MCP services have already been developed and deployed. However, the client-server integration architecture inherent in MCP may expand the attack surface against LLM Agent systems, introducing new vulnerabilities that allow attackers to exploit by designing malicious MCP servers. In th…

This https://arxiv.org/abs/2504.13865 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csHC_…

A Survey on (M)LLM-Based GUI Agents
Graphical User Interface (GUI) Agents have emerged as a transformative paradigm in human-computer interaction, evolving from rule-based automation scripts to sophisticated AI-driven systems capable of understanding and executing complex interface operations. This survey provides a comprehensive examination of the rapidly advancing field of LLM-based GUI Agents, systematically analyzing their architectural foundations, technical components, and evaluation methodologies. We identify and analyze f…

This https://arxiv.org/abs/2411.18065 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csAR_…

FlexiBit: Fully Flexible Precision Bit-parallel Accelerator Architecture for Arbitrary Mixed Precision AI
Recent research has shown that large language models (LLMs) can utilize low-precision floating point (FP) quantization to deliver high efficiency while maintaining original model accuracy. In particular, recent works have shown the effectiveness of non-power-of-two precisions, such as FP6 and FP5, and diverse sensitivity to low-precision arithmetic of LLM layers, which motivates mixed precision arithmetic including non-power-of-two precisions in LLMs. Although low-precision algorithmically lead…

Interesting piece on the #Arctic #SeaIce and it's (improving) representation in global climate models by Gavin Schmidt over on @realclimate, I'm not entirely sure all of the improvement is for the right reasons, there are probably compensating biases, it would be interesting to see if the models get the regional signal correct in different parts of the Arctic, but overall I agree with this piece, there have been some big steps forward in global models, the regional level is often not very well represented though.
https://www.realclimate.org/index.php/archives/2025/05/predicted-arctic-sea-ice-trends-over-time/

This https://arxiv.org/abs/2406.05820 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_grqc_…

Thermodynamic phase description of charged 4D Gauss-Bonnet black holes in quantum regime
Glavan and Lin [Phys. Rev. Lett. 124, 081301 (2020)] ave argued for a 4D theory of Gauss-Bonnet (GB)gravity. This model predicts significant contributions of the GB coupling parameter $α$ to gravitational dynamics, while circumventing the Lovelock theorem and avoiding Ostrogradsky instability. As a powerful competitor to general relativity (GR), the model has been examined on various phenomenological grounds. Here, we employ a technique from information geometry to analyze the thermodynamic ph…

Series B, Episode 03 - Weapon
ORAC: No. It was a priority one - for automatic relay to the senior officer present.
BLAKE: Orac, I want more information. I want to know everything there is to know about this man Coser; I want to know how he got out of the base; and I want to know what IMIPAK is.
https://blake.torpidi…

Blakes7Bot | Search
Companion site to the Blakes7Bot Twitter, Mastodon, and Bluesky accounts. Automated lines of Blakes 7 dialog. Includes a script search facility.

This https://arxiv.org/abs/2503.09492 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csIR_…

Learning Cascade Ranking as One Network
Cascade Ranking is a prevalent architecture in large-scale top-k selection systems like recommendation and advertising platforms. Traditional training methods focus on single-stage optimization, neglecting interactions between stages. Recent advances have introduced interaction-aware training paradigms, but still struggle to 1) align training objectives with the goal of the entire cascade ranking (i.e., end-to-end recall of ground-truth items) and 2) learn effective collaboration patterns for d…

Beyond the Protocol: Unveiling Attack Vectors in the Model Context Protocol Ecosystem
Hao Song, Yiming Shen, Wenxuan Luo, Leixin Guo, Ting Chen, Jiashui Wang, Beibei Li, Xiaosong Zhang, Jiachi Chen
https://arxiv.org/abs/2506.02040

Beyond the Protocol: Unveiling Attack Vectors in the Model Context Protocol Ecosystem
The Model Context Protocol (MCP) is an emerging standard designed to enable seamless interaction between Large Language Model (LLM) applications and external tools or resources. Within a short period, thousands of MCP services have already been developed and deployed. However, the client-server integration architecture inherent in MCP may expand the attack surface against LLM Agent systems, introducing new vulnerabilities that allow attackers to exploit by designing malicious MCP servers. In th…