Using the method of auxiliary coordinates, we derive a training algorithm that works by alternating steps that train an auxiliary embeddingwith steps that train the mapping.

Neural Information Processing Systems http://nips.cc/

Descriptions of complex nominal or verbal systems make use of inflectional classes. Inflectional classes bring together nouns which have similar stem changes and use similar exponents in their paradigms. Although inflectional classes can be very useful for language teaching as well as for setting up finite state morphological systems, it is unclear whether inflectional classes are cognitively real, in the sense that native speakers would need to discover these classes in order to learn how to properly inflect the nouns of their language. This study investigates whether the Discriminative Lexicon Model (DLM) can understand and produce Finnish inflected nouns without setting up inflectional classes, using a dataset with 55,271 inflected nouns of 2000 high-frequency Finnish nouns from 49 inflectional classes. Several DLM comprehension and production models were set up. Some models were not informed about frequency of use, and provide insight into learnability with infinite exposure (endstate learning). Other models were set up from a usage based perspective, and were trained with token frequencies being taken into consideration (frequency-informed learning). On training data, models performed with very high accuracies. For held-out test data, accuracies decreased, as expected, but remained acceptable. Across most models, performance increased for inflectional classes with more types, more lower-frequency words, and more hapax legomena, mirroring the productivity of the inflectional classes. The model struggles more with novel forms of unproductive and less productive classes, and performs far better for unseen forms belonging to productive classes. However, for usage-based production models, frequency was the dominant predictor of model performance, and correlations with measures of productivity were tenuous or absent.

Textual descriptions for multimodal inputs entail recurrent refinement of queries to produce relevant output images. Despite efforts to address challenges such as scaling model size and data volume, the cost associated with pre-training and fine-tuning remains substantial. However, pre-trained large multimodal models (LMMs) encounter a modality gap, characterised by a misalignment between textual and visual representations within a common embedding space. Although fine-tuning can potentially mitigate this gap, it is typically expensive and impractical due to the requirement for extensive domain-driven data. To overcome this challenge, we propose a lightweight training-free approach utilising Retrieval-Augmented Generation (RAG) to extend across the modality using a linear mapping, which can be computed efficiently. During inference, this mapping is applied to images embedded by an LMM enabling retrieval of closest textual descriptions from the training set. These textual descriptions, in conjunction with an instruction, cater as an input prompt for the language model to generate new textual descriptions. In addition, we introduce an iterative technique for distilling the mapping by generating synthetic descriptions via the language model facilitating optimisation for standard utilised image description measures. Experimental results on two benchmark multimodal datasets demonstrate significant improvements.

Large language models are computationally expensive due to their deep structures. Prior research has shown that intermediate layers contain sufficient information to generate accurate answers, leading to the development of early-exit algorithms that reduce inference costs by terminating computation at earlier layers. However, these methods often suffer from poor performance due to misalignment between intermediate and output layer representations that lead to decoding inaccuracy. To address these challenges, we propose SPADE (SPace Alignment DEcoding), a novel decoding method that aligns intermediate layer representations with the output layer by propagating a minimally reduced sequence consisting of only the start token and the answer token. We further optimize the early-exit decision-making process by training a linear approximation of SPADE that computes entropy-based confidence metrics. Putting them together, we create a hybrid early-exit algorithm that monitors confidence levels and stops inference at intermediate layers while using SPADE to generate high-quality outputs. This approach significantly reduces inference costs without compromising accuracy, offering a scalable and efficient solution for deploying large language models in real-world applications.