Query routing, the task to route user queries to different large language model (LLM) endpoints, can be considered as a text classification problem. However, out-of-distribution queries must be handled properly, as those could be about unrelated domains, queries in other languages, or even contain unsafe text. Here, we thus study a guarded query routing problem, for which we first introduce the Guarded Query Routing Benchmark (GQR-Bench, released as Python package gqr), covers three exemplary target domains (law, finance, and healthcare), and seven datasets to test robustness against out-of-distribution queries. We then use GQR-Bench to contrast the effectiveness and efficiency of LLM-based routing mechanisms (GPT-4o-mini, Llama-3.2-3B, and Llama-3.1-8B), standard LLM-based guardrail approaches (LlamaGuard and NVIDIA NeMo Guardrails), continuous bag-of-words classifiers (WideMLP, fastText), and traditional machine learning models (SVM, XGBoost). Our results show that WideMLP, enhanced with out-of-domain detection capabilities, yields the best trade-off between accuracy (88%) and speed (<4ms). The embedding-based fastText excels at speed (<1ms) with acceptable accuracy (80%), whereas LLMs yield the highest accuracy (91%) but are comparatively slow (62ms for local Llama-3.1:8B and 669ms for remote GPT-4o-mini calls). Our findings challenge the automatic reliance on LLMs for (guarded) query routing and provide concrete recommendations for practical applications. Source code is available: https://github.com/williambrach/gqr.

Cryptocurrency arbitrage, a riskless trading strategy, can yield profits but requires swift execution due to volatile opportunities that vanish rapidly. Utilizing arbitrage bots for algorithmic trading is essential for immediate trade execution across exchanges like Binance and Bybit. This paper implements such a system focusing on BTCUSDT and ETHUSDT pairs. Integrating Machine Learning (ML) aims to predict arbitrage occurrences in advance for faster trade execution, a tactic many traders overlook. Logistic Regression, Random Forest, Support Vector Machine, and Multilayer Perceptron models are implemented. Adding ML principles required the collection of a dataset with historical prices of the observed cryptocurrency pairs for various time intervals, on which we trained the model. Afterward, the model was evaluated in a live-trading environment. Results show Random Forest predicting exploitable arbitrage intervals ahead for Ether, with ML models more effective during less volatile periods. However, careful consideration is needed as predictions may not always align with market realities, leading to mixed trading outcomes. Furthermore, the training led to a model that can predict the occurrence of arbitrage; however, classifying the calculations even more carefully than in reality, resulting in a partially profitable or partially lossy trading strategy depending on the time of day and the current market stage.

Blockchain is a technology that is often used to share data and assets. However, in the decentralized ecosystem, blockchain-based systems can be utilized to share information and assets without the traditional barriers associated with solo responsibility, e.g., multi-sig wallets. This paper describes an innovative approach to blockchain networks based on a non-fungible token that behaves as an account (NFTAA). The key novelty of this article is using NFTAA to leverage the unique properties of NFTs to manage your ownership better and effectively isolate them to improve the security, transparency, and even interoperability possibilities. Additionally, the account-based solution gives us the ability and flexibility to cover regular use cases such a staking and liquid equities, but also practical composability. This article offers a simple implementation, which allows developers and researcher to choose the best solution for their needs in demand of abstract representation in any use case.

The first blockchain projects were not built with interoperability in mind. This idea became more popular later as development progressed, and there was a need for the ability to share assets and messages with other blockchain networks. Some of the blockchains that were not made with an interoperability focus started realizing that it was where the future was headed and implemented interoperability that existed, or they created their own way of communicating with other networks. To this day, many protocols are being built that enhance interoperability. Blockchain interoperability refers to the ability of different blockchain systems to work together and communicate with each other. This is an important concept for the future of blockchain technology, as it enables different blockchain networks to exchange information and assets seamlessly. This paper focuses on enhancing the interoperability between Paraverse and the specific network called Polkadot. This is achieved by using the network-specific native protocol for interoperability, called XCMP. The results show that not only is ParaSpell SDK a useful contribution to the Polkadot ecosystem, but it also makes interoperable transactions easier and more user-friendly.