English
Related papers

Related papers: Weight Tying Biases Token Embeddings Towards the O…

200 papers

Embedding layers in transformer-based NLP models typically account for the largest share of model parameters, scaling with vocabulary size but not yielding performance gains proportional to scale. We propose an alternative approach in which…

Computation and Language · Computer Science 2025-05-06 Henry Ndubuaku , Mouad Talhi

Models trained on different datasets can be merged by a weighted-averaging of their parameters, but why does it work and when can it fail? Here, we connect the inaccuracy of weighted-averaging to mismatches in the gradients and propose a…

Machine Learning · Computer Science 2024-08-26 Nico Daheim , Thomas Möllenhoff , Edoardo Maria Ponti , Iryna Gurevych , Mohammad Emtiyaz Khan

Machine learning algorithms are optimized to model statistical properties of the training data. If the input data reflects stereotypes and biases of the broader society, then the output of the learning algorithm also captures these…

Computation and Language · Computer Science 2016-06-21 Tolga Bolukbasi , Kai-Wei Chang , James Zou , Venkatesh Saligrama , Adam Kalai

Results in interpretability suggest that large vision and language models learn implicit linear encodings when models are biased by in-context prompting. However, the existence of similar linear representations in more general adaptation…

Machine Learning · Computer Science 2025-12-18 Darrin O' Brien , Dhikshith Gajulapalli , Eric Xia

Multi-task learning improves generalization, but when does it reduce the model capacity required to learn? We provide a systematic study of how joint training affects the learning transition, the minimum model size at which a task can be…

Machine Learning · Computer Science 2026-02-03 Csaba Both , Benjamin Hoover , Hendrik Strobelt , Dmitry Krotov , Daniel Karl I. Weidele , Mauro Martino , Nima Dehmamy

Embedding-based neural topic models could explicitly represent words and topics by embedding them to a homogeneous feature space, which shows higher interpretability. However, there are no explicit constraints for the training of…

Computation and Language · Computer Science 2022-06-17 Wei Shao , Lei Huang , Shuqi Liu , Shihua Ma , Linqi Song

Transformer models learn to encode and decode an input text, and produce contextual token embeddings as a side-effect. The mapping from language into the embedding space maps words expressing similar concepts onto points that are close in…

Computation and Language · Computer Science 2025-09-03 Vivi Nastase , Paola Merlo

Neural language models learn word representations, or embeddings, that capture rich linguistic and conceptual information. Here we investigate the embeddings learned by neural machine translation models, a recently-developed class of neural…

Computation and Language · Computer Science 2015-04-06 Felix Hill , Kyunghyun Cho , Sebastien Jean , Coline Devin , Yoshua Bengio

The embedding layers transforming input words into real vectors are the key components of deep neural networks used in natural language processing. However, when the vocabulary is large, the corresponding weight matrices can be enormous,…

Computation and Language · Computer Science 2020-02-20 Oleksii Hrinchuk , Valentin Khrulkov , Leyla Mirvakhabova , Elena Orlova , Ivan Oseledets

Large language models based on transformers have achieved great empirical successes. However, as they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable.…

Machine Learning · Statistics 2023-11-08 Alberto Bietti , Vivien Cabannes , Diane Bouchacourt , Herve Jegou , Leon Bottou

Multi-task post-training of large language models (LLMs) is typically performed by mixing datasets from different tasks and optimizing them jointly. This approach implicitly assumes that all tasks contribute gradients of similar magnitudes;…

Text embedding is becoming an increasingly popular AI methodology, especially among businesses, yet the potential of text embedding models to be biased is not well understood. This paper examines the degree to which a selection of popular…

Artificial Intelligence · Computer Science 2024-06-19 Vasyl Rakivnenko , Nestor Maslej , Jessica Cervi , Volodymyr Zhukov

Embeddings are a powerful way to enrich data-driven machine learning models with the world knowledge of large language models (LLMs). Yet, there is limited evidence on how to design effective LLM-based embedding pipelines for tabular…

Machine Learning · Computer Science 2026-03-19 Oksana Kolomenko , Ricardo Knauer , Erik Rodner

Providing high-quality feedback to Large Language Models (LLMs) on a diverse training distribution can be difficult and expensive, and providing feedback only on a narrow distribution can result in unintended generalizations. To better…

Computation and Language · Computer Science 2026-03-02 Constanza Fierro , Fabien Roger

Large language models (LLMs) are pretrained by minimizing the cross-entropy loss for next-token prediction. In this paper, we study whether this optimization strategy can induce geometric structure in the learned model weights and context…

Optimization and Control · Mathematics 2026-05-14 Zhehang Du , Hangfeng He , Weijie Su

Masked Language Modeling (MLM) is widely used to pretrain language models. The standard random masking strategy in MLM causes the pre-trained language models (PLMs) to be biased toward high-frequency tokens. Representation learning of rare…

Computation and Language · Computer Science 2023-05-25 Linhan Zhang , Qian Chen , Wen Wang , Chong Deng , Xin Cao , Kongzhang Hao , Yuxin Jiang , Wei Wang

Text embeddings from large language models (LLMs) have achieved excellent results in tasks such as information retrieval, semantic textual similarity, etc. In this work, we show an interesting finding: when feeding a text into the LLM-based…

Computation and Language · Computer Science 2025-07-08 Zhijie Nie , Richong Zhang , Zhanyu Wu

Understanding the internal mechanisms of large language models (LLMs) remains a challenging and complex endeavor. Even fundamental questions, such as how fine-tuning affects model behavior, often require extensive empirical evaluation. In…

Task vector composition has emerged as a promising paradigm for editing pre-trained models, enabling model merging through addition and unlearning through subtraction. Fine-tuning in the tangent space of a pre-trained model (linear…

Machine Learning · Computer Science 2026-05-25 Thomas Sommariva , Francesca Morandi , Simone Calderara , Angelo Porrello

Given multiple source word embeddings learnt using diverse algorithms and lexical resources, meta word embedding learning methods attempt to learn more accurate and wide-coverage word embeddings. Prior work on meta-embedding has repeatedly…

Computation and Language · Computer Science 2022-04-27 Danushka Bollegala