Analyzing the Inner Workings of Transformers in Compositional Generalization

• URL: http://arxiv.org/abs/2502.15277v1
• Date: Fri, 21 Feb 2025 08:07:53 GMT
• Title: Analyzing the Inner Workings of Transformers in Compositional Generalization
• Authors: Ryoma Kumon, Hitomi Yanaka,
• Abstract summary: We investigate the inner workings of a Transformer model by finding an existing subnetwork that contributes to the generalization performance.<n>We find that the model depends on syntactic features to output the correct answer, but that the subnetwork with much better generalization performance than the whole model relies on a non-compositional algorithm.
• Score: 15.599899071518545
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The compositional generalization abilities of neural models have been sought after for human-like linguistic competence. The popular method to evaluate such abilities is to assess the models' input-output behavior. However, that does not reveal the internal mechanisms, and the underlying competence of such models in compositional generalization remains unclear. To address this problem, we explore the inner workings of a Transformer model by finding an existing subnetwork that contributes to the generalization performance and by performing causal analyses on how the model utilizes syntactic features. We find that the model depends on syntactic features to output the correct answer, but that the subnetwork with much better generalization performance than the whole model relies on a non-compositional algorithm in addition to the syntactic features. We also show that the subnetwork improves its generalization performance relatively slowly during the training compared to the in-distribution one, and the non-compositional solution is acquired in the early stages of the training.

Related papers

• Investigating the Role of Instruction Variety and Task Difficulty in Robotic Manipulation Tasks [50.75902473813379]
  This work introduces a comprehensive evaluation framework that systematically examines the role of instructions and inputs in the generalisation abilities of such models. The proposed framework uncovers the resilience of multimodal models to extreme instruction perturbations and their vulnerability to observational changes.
  arXiv  Detail & Related papers  (2024-07-04T14:36:49Z)
• Uncovering mesa-optimization algorithms in Transformers [61.06055590704677]
  Some autoregressive models can learn as an input sequence is processed, without undergoing any parameter changes, and without being explicitly trained to do so. We show that standard next-token prediction error minimization gives rise to a subsidiary learning algorithm that adjusts the model as new inputs are revealed. Our findings explain in-context learning as a product of autoregressive loss minimization and inform the design of new optimization-based Transformer layers.
  arXiv  Detail & Related papers  (2023-09-11T22:42:50Z)
• Bayesian Prompt Learning for Image-Language Model Generalization [64.50204877434878]
  We use the regularization ability of Bayesian methods to frame prompt learning as a variational inference problem. Our approach regularizes the prompt space, reduces overfitting to the seen prompts and improves the prompt generalization on unseen prompts. We demonstrate empirically on 15 benchmarks that Bayesian prompt learning provides an appropriate coverage of the prompt space.
  arXiv  Detail & Related papers  (2022-10-05T17:05:56Z)
• On the Generalization and Adaption Performance of Causal Models [99.64022680811281]
  Differentiable causal discovery has proposed to factorize the data generating process into a set of modules. We study the generalization and adaption performance of such modular neural causal models. Our analysis shows that the modular neural causal models outperform other models on both zero and few-shot adaptation in low data regimes.
  arXiv  Detail & Related papers  (2022-06-09T17:12:32Z)
• Amortized Inference for Causal Structure Learning [72.84105256353801]
  Learning causal structure poses a search problem that typically involves evaluating structures using a score or independence test. We train a variational inference model to predict the causal structure from observational/interventional data. Our models exhibit robust generalization capabilities under substantial distribution shift.
  arXiv  Detail & Related papers  (2022-05-25T17:37:08Z)
• Revisiting the Compositional Generalization Abilities of Neural Sequence Models [23.665350744415004]
  We focus on one-shot primitive generalization as introduced by the popular SCAN benchmark. We demonstrate that modifying the training distribution in simple and intuitive ways enables standard seq-to-seq models to achieve near-perfect generalization performance.
  arXiv  Detail & Related papers  (2022-03-14T18:03:21Z)
• Rethinking Self-Supervision Objectives for Generalizable Coherence Modeling [8.329870357145927]
  Coherence evaluation of machine generated text is one of the principal applications of coherence models that needs to be investigated. We explore training data and self-supervision objectives that result in a model that generalizes well across tasks. We show empirically that increasing the density of negative samples improves the basic model, and using a global negative queue further improves and stabilizes the model while training with hard negative samples.
  arXiv  Detail & Related papers  (2021-10-14T07:44:14Z)
• Refining Neural Networks with Compositional Explanations [31.84868477264624]
  We propose to refine a learned model by collecting human-provided compositional explanations on the models' failure cases. We demonstrate the effectiveness of the proposed approach on two text classification tasks.
  arXiv  Detail & Related papers  (2021-03-18T17:48:54Z)
• More Is More -- Narrowing the Generalization Gap by Adding Classification Heads [8.883733362171032]
  We introduce an architecture enhancement for existing neural network models based on input transformations, termed 'TransNet' Our model can be employed during training time only and then pruned for prediction, resulting in an equivalent architecture to the base model.
  arXiv  Detail & Related papers  (2021-02-09T16:30:33Z)
• Multiplicative noise and heavy tails in stochastic optimization [62.993432503309485]
  empirical optimization is central to modern machine learning, but its role in its success is still unclear. We show that it commonly arises in parameters of discrete multiplicative noise due to variance. A detailed analysis is conducted in which we describe on key factors, including recent step size, and data, all exhibit similar results on state-of-the-art neural network models.
  arXiv  Detail & Related papers  (2020-06-11T09:58:01Z)
• A Systematic Assessment of Syntactic Generalization in Neural Language Models [20.589737524626745]
  We present a systematic evaluation of the syntactic knowledge of neural language models. We find substantial differences in syntactic generalization performance by model architecture. Our results also reveal a dissociation between perplexity and syntactic generalization performance.
  arXiv  Detail & Related papers  (2020-05-07T18:35:25Z)
• Rethinking Generalization of Neural Models: A Named Entity Recognition Case Study [81.11161697133095]
  We take the NER task as a testbed to analyze the generalization behavior of existing models from different perspectives. Experiments with in-depth analyses diagnose the bottleneck of existing neural NER models. As a by-product of this paper, we have open-sourced a project that involves a comprehensive summary of recent NER papers.
  arXiv  Detail & Related papers  (2020-01-12T04:33:53Z)

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.