arXiv subjects

Andrew Pulver

Publications and source records attributed to Andrew Pulver.

3 recordsLinked to original sources

arXiv2025

PerQ: Efficient Evaluation of Multilingual Text Personalization Quality

Since no metrics are available to evaluate specific aspects of a text, such as its personalization quality, the researchers often rely solely on large language models to meta-evaluate such texts. Due to internal biases of individual language models, it is recommended to use multiple of them for combined evaluation, which directly increases costs of such meta-evaluation. In this paper, a computationally efficient method for evaluation of personalization quality of a given text (generated by a language model) is introduced, called PerQ. A case study of comparison of generation capabilities of large and small language models shows the usability of the proposed metric in research, effectively reducing the waste of resources.

cs.CL↗

arXiv2021

On Problems Dual to Unification: The String-Rewriting Case

In this paper, we investigate problems which are dual to the unification problem, namely the Fixed Point (FP) problem, Common Term (CT) problem and the Common Equation (CE) problem for string rewriting systems. Our main motivation is computing fixed points in systems, such as loop invariants in programming languages. We show that the fixed point (FP) problem is reducible to the common term problem. Our new results are: (i) the fixed point problem is undecidable for finite convergent string rewriting systems whereas it is decidable in polynomial time for finite, convergent and dwindling string rewriting systems, (ii) the common term problem is undecidable for the class of dwindling string rewriting systems, and (iii) for the class of finite, monadic and convergent systems, the common equation problem is decidable in polynomial time but for the class of dwindling string rewriting systems, common equation problem is undecidable.

cs.LO↗

arXiv2017

LSTM with Working Memory

Previous RNN architectures have largely been superseded by LSTM, or "Long Short-Term Memory". Since its introduction, there have been many variations on this simple design. However, it is still widely used and we are not aware of a gated-RNN architecture that outperforms LSTM in a broad sense while still being as simple and efficient. In this paper we propose a modified LSTM-like architecture. Our architecture is still simple and achieves better performance on the tasks that we tested on. We also introduce a new RNN performance benchmark that uses the handwritten digits and stresses several important network capabilities.

cs.NE↗