Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models

• URL: http://arxiv.org/abs/2502.09863v3
• Date: Thu, 16 Oct 2025 19:53:18 GMT
• Title: Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like Models
• Abstract summary: Self-supervised word embedding algorithms such as word2vec provide a minimal setting for studying representation learning in language modeling.<n>We show that both the training dynamics and the final performance on downstream tasks are empirically very similar to those of word2vec.
• Score: 5.015875080287129
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Self-supervised word embedding algorithms such as word2vec provide a minimal setting for studying representation learning in language modeling. We examine the quartic Taylor approximation of the word2vec loss around the origin, and we show that both the resulting training dynamics and the final performance on downstream tasks are empirically very similar to those of word2vec. Our main contribution is to analytically solve for both the gradient flow training dynamics and the final word embeddings in terms of only the corpus statistics and training hyperparameters. The solutions reveal that these models learn orthogonal linear subspaces one at a time, each one incrementing the effective rank of the embeddings until model capacity is saturated. Training on Wikipedia, we find that each of the top linear subspaces represents an interpretable topic-level concept. Finally, we apply our theory to describe how linear representations of more abstract semantic concepts emerge during training; these can be used to complete analogies via vector addition.

Related papers

• Implicit Bias of SGD in Multivariate ReLU Networks: Effective Width Collapse [16.78534503577613]
  We show that the input weights and biases align along finitely many directions, leading to an effective width collapse.<n>We establish a non-redundancy property of the learned representation by proving that each learned direction induces a unique ternary activation pattern on the training data.
  arXiv  Detail & Related papers  (2026-07-03T22:03:13Z)
• Vocabulary embeddings organize linguistic structure early in language model training [3.2661767443292646]
  Large language models (LLMs) work by manipulating the geometry of input embedding vectors over multiple layers.<n>Here, we ask: how are the input vocabulary representations of language models structured, and how does this structure evolve over training?<n>We run a suite of experiments that correlate the geometric structure of the input embeddings and output embeddings of two open-source models with semantic, syntactic, and frequency-based metrics over the course of training.
  arXiv  Detail & Related papers  (2025-10-08T23:26:22Z)
• Grokking ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior [25.975757048963413]
  Post-hoc interpretability methods typically attribute a model's behavior to its components, data, or training trajectory in isolation.<n>We present ExPLAIND, a unified framework that integrates all three perspectives.
  arXiv  Detail & Related papers  (2025-05-26T14:53:11Z)
• Large Concept Models: Language Modeling in a Sentence Representation Space [62.73366944266477]
  We present an attempt at an architecture which operates on an explicit higher-level semantic representation, which we name a concept.<n> Concepts are language- and modality-agnostic and represent a higher level idea or action in a flow.<n>We show that our model exhibits impressive zero-shot generalization performance to many languages.
  arXiv  Detail & Related papers  (2024-12-11T23:36:20Z)
• Bilinear Convolution Decomposition for Causal RL Interpretability [0.0]
  Efforts to interpret reinforcement learning (RL) models often rely on high-level techniques such as attribution or probing.<n>This work proposes replacing nonlinearities in convolutional neural networks (ConvNets) with bilinear variants, to produce a class of models for which these limitations can be addressed.<n>We show bilinear model variants perform comparably in model-free reinforcement learning settings, and give a side by side comparison on ProcGen environments.
  arXiv  Detail & Related papers  (2024-12-01T19:32:04Z)
• Graph-Dictionary Signal Model for Sparse Representations of Multivariate Data [49.77103348208835]
  We define a novel Graph-Dictionary signal model, where a finite set of graphs characterizes relationships in data distribution through a weighted sum of their Laplacians. We propose a framework to infer the graph dictionary representation from observed data, along with a bilinear generalization of the primal-dual splitting algorithm to solve the learning problem. We exploit graph-dictionary representations in a motor imagery decoding task on brain activity data, where we classify imagined motion better than standard methods.
  arXiv  Detail & Related papers  (2024-11-08T17:40:43Z)
• State-space models can learn in-context by gradient descent [1.3087858009942543]
  We show that state-space models can perform gradient-based learning and use it for in-context learning in much the same way as transformers.<n>Specifically, we prove that a single structured state-space model layer, augmented with multiplicative input and output gating, can reproduce the outputs of an implicit linear model.<n>We also provide novel insights into the relationship between state-space models and linear self-attention, and their ability to learn in-context.
  arXiv  Detail & Related papers  (2024-10-15T15:22:38Z)
• Representations as Language: An Information-Theoretic Framework for Interpretability [7.2129390689756185]
  Large scale neural models show impressive performance across a wide array of linguistic tasks. Despite this they remain, largely, black-boxes, inducing vector-representations of their input that prove difficult to interpret. We introduce a novel approach to interpretability that looks at the mapping a model learns from sentences to representations as a kind of language in its own right.
  arXiv  Detail & Related papers  (2024-06-04T16:14:00Z)
• Action Model Learning with Guarantees [5.524804393257921]
  We develop a theory for action model learning based on version spaces that interprets the task as search for hypothesis that are consistent with the learning examples. Our theoretical findings are instantiated in an online algorithm that maintains a compact representation of all solutions of the problem.
  arXiv  Detail & Related papers  (2024-04-15T10:01:43Z)
• On the Origins of Linear Representations in Large Language Models [51.88404605700344]
  We introduce a simple latent variable model to formalize the concept dynamics of the next token prediction. Experiments show that linear representations emerge when learning from data matching the latent variable model. We additionally confirm some predictions of the theory using the LLaMA-2 large language model.
  arXiv  Detail & Related papers  (2024-03-06T17:17:36Z)
• In-Context Language Learning: Architectures and Algorithms [73.93205821154605]
  We study ICL through the lens of a new family of model problems we term in context language learning (ICLL) We evaluate a diverse set of neural sequence models on regular ICLL tasks.
  arXiv  Detail & Related papers  (2024-01-23T18:59:21Z)
• Transparency at the Source: Evaluating and Interpreting Language Models With Access to the True Distribution [4.01799362940916]
  We present a setup for training, evaluating and interpreting neural language models, that uses artificial, language-like data. The data is generated using a massive probabilistic grammar, that is itself derived from a large natural language corpus. With access to the underlying true source, our results show striking differences and outcomes in learning dynamics between different classes of words.
  arXiv  Detail & Related papers  (2023-10-23T12:03:01Z)
• Meaning Representations from Trajectories in Autoregressive Models [106.63181745054571]
  We propose to extract meaning representations from autoregressive language models by considering the distribution of all possible trajectories extending an input text. This strategy is prompt-free, does not require fine-tuning, and is applicable to any pre-trained autoregressive model. We empirically show that the representations obtained from large models align well with human annotations, outperform other zero-shot and prompt-free methods on semantic similarity tasks, and can be used to solve more complex entailment and containment tasks that standard embeddings cannot handle.
  arXiv  Detail & Related papers  (2023-10-23T04:35:58Z)
• Opening the Black Box: Analyzing Attention Weights and Hidden States in Pre-trained Language Models for Non-language Tasks [0.8889304968879164]
  We apply a pre-trained language model to constrained arithmetic problems with hierarchical structure, to analyze their attention weight scores and hidden states. The investigation reveals promising results, with the model addressing hierarchical problems in a moderately structured manner, similar to human problem-solving strategies. The attention analysis allows us to hypothesize that the model can generalize to longer sequences in ListOps dataset, a conclusion later confirmed through testing on sequences longer than those in the training set.
  arXiv  Detail & Related papers  (2023-06-21T11:48:07Z)
• Latent Traversals in Generative Models as Potential Flows [113.4232528843775]
  We propose to model latent structures with a learned dynamic potential landscape. Inspired by physics, optimal transport, and neuroscience, these potential landscapes are learned as physically realistic partial differential equations. Our method achieves both more qualitatively and quantitatively disentangled trajectories than state-of-the-art baselines.
  arXiv  Detail & Related papers  (2023-04-25T15:53:45Z)
• Training Trajectories of Language Models Across Scales [99.38721327771208]
  Scaling up language models has led to unprecedented performance gains. How do language models of different sizes learn during pre-training? Why do larger language models demonstrate more desirable behaviors?
  arXiv  Detail & Related papers  (2022-12-19T19:16:29Z)
• Low-Rank Constraints for Fast Inference in Structured Models [110.38427965904266]
  This work demonstrates a simple approach to reduce the computational and memory complexity of a large class of structured models. Experiments with neural parameterized structured models for language modeling, polyphonic music modeling, unsupervised grammar induction, and video modeling show that our approach matches the accuracy of standard models at large state spaces.
  arXiv  Detail & Related papers  (2022-01-08T00:47:50Z)
• Implicit Representations of Meaning in Neural Language Models [31.71898809435222]
  We identify contextual word representations that function as models of entities and situations as they evolve throughout a discourse. Our results indicate that prediction in pretrained neural language models is supported, at least in part, by dynamic representations of meaning and implicit simulation of entity state.
  arXiv  Detail & Related papers  (2021-06-01T19:23:20Z)
• TCL: Transformer-based Dynamic Graph Modelling via Contrastive Learning [87.38675639186405]
  We propose a novel graph neural network approach, called TCL, which deals with the dynamically-evolving graph in a continuous-time fashion. To the best of our knowledge, this is the first attempt to apply contrastive learning to representation learning on dynamic graphs.
  arXiv  Detail & Related papers  (2021-05-17T15:33:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.