The recent success of large language models (LLMs) has been predominantly driven by curating the training dataset composition, scaling of model architectures and dataset sizes and advancements in pretraining objectives, leaving tokenizer influence as a blind spot.Shedding light on this underexplored area, we conduct a comprehensive study on the influence of tokenizer choice on LLM downstream performance by training 24 mono- and multilingual LLMs at a 2.6B parameter scale, ablating different tokenizer algorithms and parameterizations. Our studies highlight that the tokenizer choice can significantly impact the model’s downstream performance and training costs. In particular, we find that the common tokenizer evaluation metrics fertility and parity are not always predictive of model downstream performance, rendering these metrics a questionable proxy for the model’s downstream performance. Furthermore, we show that multilingual tokenizers trained on the five most frequent European languages require vocabulary size increases of factor three in comparison to English. While English-centric tokenizers have been applied to the training of multi-lingual LLMs in the past, we find that this approach results in a severe downstream performance degradation and additional training costs of up to 68%, due to an inefficient tokenization vocabulary.

This paper describes the development of a structured document collection containing user-generated text and numerical metadata for exploring the exploitation of metadata in information retrieval (IR). The collection consists of more than 61,000 documents extracted from YouTube video pages on basketball in general and NBA (National Basketball Association) in particular, together with a set of 40 topics and their relevance judgements. In addition, a collection of nearly 250,000 user profiles related to the NBA collection is available. Several baseline IR experiments report the effect of using video-associated metadata on retrieval effectiveness. The results surprisingly show that searching the videos titles only performs significantly better than searching additional metadata text fields of the videos such as the tags or the description.

LRs remain expensive to create and thus rare relative to demand across languages and technology types. The accidental re-creation of an LR that already exists is a nearly unforgivable waste of scarce resources that is unfortunately not so easy to avoid. The number of catalogs the HLT researcher must search, with their different formats, make it possible to overlook an existing resource. This paper sketches the sources of this problem and outlines a proposal to rectify along with a new vision of LR cataloging that will to facilitates the documentation and exploitation of a much wider range of LRs than previously considered.