Tootfinder

This https://arxiv.org/abs/2505.22767 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csHC_…

In Dialogue with Intelligence: Rethinking Large Language Models as Collective Knowledge
Large Language Models (LLMs) are typically analysed through architectural, behavioural, or training-data lenses. This article offers a theoretical and experiential re-framing: LLMs as dynamic instantiations of Collective human Knowledge (CK), where intelligence is evoked through dialogue rather than stored statically. Drawing on concepts from neuroscience and AI, and grounded in sustained interaction with ChatGPT-4, I examine emergent dialogue patterns, the implications of fine-tuning, and the …

Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMs
Chenjun Xu, Bingbing Wen, Bin Han, Robert Wolfe, Lucy Lu Wang, Bill Howe
https://arxiv.org/abs/2506.00582

Do Language Models Mirror Human Confidence? Exploring Psychological Insights to Address Overconfidence in LLMs
Psychology research has shown that humans are poor at estimating their performance on tasks, tending towards underconfidence on easy tasks and overconfidence on difficult tasks. We examine three LLMs, Llama-3-70B-instruct, Claude-3-Sonnet, and GPT-4o, on a range of QA tasks of varying difficulty, and show that models exhibit subtle differences from human patterns of overconfidence: less sensitive to task difficulty, and when prompted to answer based on different personas -- e.g., expert vs laym…

We have been public about not using Go as the bootstrap language. We started with a compiler in C for a number of reasons, mostly for ease of bootstrapping but also to avoid making a compiler-optimized language instead of a general purpose one. Once the language was designed, we ported the code to Go for ease of maintenance and development.
Well here is a variant of that pattern: the TypeScript compiler is also being ported to Go: https://www.youtube.com/watch?v=pNlq-EVld70
The video is worth watching for the reasons why. Turns out a general purpose language is indeed a good compiler language.

On Reconfigurable Bisimulation, with an Application to the Distributed Synthesis Problem
Yehia Abd Alrahman, Nir Piterman
https://arxiv.org/abs/2505.21672 …

On Reconfigurable Bisimulation, with an Application to the Distributed Synthesis Problem
We consider the problem of distributing a centralised transition system to a set of asynchronous agents recognising the same language. Existing solutions are either manual or involve a huge explosion in the number of states from the centralised system. The difficulty arises from the need to keep a rigid communication scheme, specifying a fixed mapping from events to those who can participate in them. Thus, individual agents need to memorise seen events and their order to dynamically compare the…

Understanding the Information Propagation Effects of Communication Topologies in LLM-based Multi-Agent Systems
Xu Shen, Yixin Liu, Yiwei Dai, Yili Wang, Rui Miao, Yue Tan, Shirui Pan, Xin Wang
https://arxiv.org/abs/2505.23352

Understanding the Information Propagation Effects of Communication Topologies in LLM-based Multi-Agent Systems
The communication topology in large language model-based multi-agent systems fundamentally governs inter-agent collaboration patterns, critically shaping both the efficiency and effectiveness of collective decision-making. While recent studies for communication topology automated design tend to construct sparse structures for efficiency, they often overlook why and when sparse and dense topologies help or hinder collaboration. In this paper, we present a causal framework to analyze how agent ou…

With the advent of ELIZA, Joseph Weizenbaum's first psychotherapist chatbot, NLP took another major step with pattern-based substitution algorithms based on simple regular expressions.
Weizenbaum, Joseph (1966). ELIZA—a computer program for the study of natural language communication between man and machine. Com. of the ACM. 9: 36–45.

This https://arxiv.org/abs/2405.07671 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csFL_…

Constructing a BPE Tokenization DFA
Many natural language processing systems operate over tokenizations of text to address the open-vocabulary problem. In this paper, we give and analyze an algorithm for the efficient construction of deterministic finite automata (DFA) designed to operate directly on tokenizations produced by the popular byte pair encoding (BPE) technique. This makes it possible to apply many existing techniques and algorithms to the tokenized case, such as pattern matching, equivalence checking of tokenization d…

Towards Efficient Key-Value Cache Management for Prefix Prefilling in LLM Inference
Yue Zhu, Hao Yu, Chen Wang, Zhuoran Liu, Eun Kyung Lee
https://arxiv.org/abs/2505.21919

Towards Efficient Key-Value Cache Management for Prefix Prefilling in LLM Inference
The increasing adoption of large language models (LLMs) with extended context windows necessitates efficient Key-Value Cache (KVC) management to optimize inference performance. Inference workloads like Retrieval-Augmented Generation (RAG) and agents exhibit high cache reusability, making efficient caching critical to reducing redundancy and improving speed. We analyze real-world KVC access patterns using publicly available traces and evaluate commercial key-value stores like Redis and state-of-…

I'm interested in making my own visualizations of weather radar data to display patterns in a region of the US (covered by multiple stations) over longish periods of time (e.g., a month).
Think something like this year of US radar, but smaller & shorter in my case: https://www.youtube.com/watch?v=FPuDPxK88g0
I'm familiar with mapping and animation in Python/R, but not weather radar data. I've been looking for tutorials, but haven't been able to track any down yet. Pointers very much welcome

Performance of Confidential Computing GPUs
Antonio Mart\'inez Ibarra, Julian James Stephen, Aurora Gonz\'alez Vidal, K. R. Jayaram, Antonio Fernando Skarmeta G\'omez
https://arxiv.org/abs/2505.16501

Performance of Confidential Computing GPUs
This work examines latency, throughput, and other metrics when performing inference on confidential GPUs. We explore different traffic patterns and scheduling strategies using a single Virtual Machine with one NVIDIA H100 GPU, to perform relaxed batch inferences on multiple Large Language Models (LLMs), operating under the constraint of swapping models in and out of memory, which necessitates efficient control. The experiments simulate diverse real-world scenarios by varying parameters such as …