Challenge: Existing models for deep transformers are able to combine word meanings into phrase meanings, but they lack a clear understanding of how they handle complex linguistic inputs.
Approach: They propose to analyze phrasal representations in pre-trained transformers to determine whether they reflect sophisticated composition of phrase meaning.
Outcome: The proposed models are able to combine meaning units into larger units, a phenomenon known as composition, and reflects human understanding of meaning.

Similar Papers

On the Interplay Between Fine-tuning and Composition in Transformers (2021.findings-acl)

Copied to clipboard

Challenge: Pre-trained transformer language models have shown remarkable performance on a variety of NLP tasks.
Approach: They propose to fine-tune transformer language models on a paraphrase and sentiment task and analyze their results to determine whether they benefit compositionality.
Outcome: The proposed model performance on a paraphrase and sentiment task is compared with pre-trained models on lexical-level representations.
How do Transformer Embeddings Represent Compositions? A Functional Analysis (2025.findings-acl)

Copied to clipboard

Challenge: Despite the popularity of transformer-based models, little is known about how they represent compound words and whether they are compositional.
Approach: They evaluate compositionality in mistral, OpenAI Large, and Google embedding models and compare them with BERT.
Outcome: The proposed models perform best in addition, multiplication, dilation, regression, and the classic vector addition model performs almost as well as any other model.
The Impact of Depth on Compositional Generalization in Transformer Language Models (2024.naacl-long)

Copied to clipboard

Challenge: In this paper, we test the hypothesis that deeper transformers generalize more compositionally.
Approach: They propose to add layers to transformers to generalize more compositionally . they propose to fine-tune the models so that the total number of parameters is constant .
Outcome: The proposed model generalizes more compositionally than shallower models, but returns diminish . the proposed model can be made shallower without sacrificing performance .
Making Transformers Solve Compositional Tasks (2022.acl-long)

Copied to clipboard

Challenge: Several studies have reported the inability of Transformer models to generalize compositionally . a key aspect of natural language is the ability to learn basic primitives .
Approach: They propose to use Transformers to generalize compositionally in a large range of tasks . they find that Transformers generalize significantly better than previous models .
Outcome: The proposed models generalize compositionally significantly better than previous models . a set of 12 datasets shows that the proposed models can be improved .
Quantifying Compositionality of Classic and State-of-the-Art Embeddings (2025.findings-emnlp)

Copied to clipboard

Challenge: Static word embeddings make strong claims about compositionality, but the SOTA generative models go too far in the other direction.
Approach: a new study evaluates the compositionality of word embeddings by canonical correlation analysis . strong compositional signals are observed in later training stages across data modalities .
Outcome: a new evaluation of compositional models shows that they exploit access meanings when justified . strong compositional signals are observed in later training stages and in deeper layers of the transformer-based model before a decline at the top layer.
Word Representation Learning in Multimodal Pre-Trained Transformers: An Intrinsic Evaluation (2021.tacl-1)

Copied to clipboard

Challenge: Existing models for linguistic representations of words are based on information extracted from large text corpora, and the sensory-motor experiences humans have with the world play an important role in determining word meaning.
Approach: They propose to use contextualized word representations to learn semantic representations of words that align with human semantic intuitions.
Outcome: The proposed models are shown to be more efficient on concrete word pairs than on abstract ones.
Analyzing the Inner Workings of Transformers in Compositional Generalization (2025.naacl-long)

Copied to clipboard

Challenge: Existing studies on compositional generalization abilities of neural models have focused on benchmarks, but the results do not reflect the underlying competence of the model.
Approach: They propose to find an existing subnetwork that contributes to the generalization performance and perform causal analyses on how the model utilizes syntactic features.
Outcome: The proposed model relies on syntactic features but the subnetwork with better generalization performance relies mainly on a non-compositional algorithm .
Explaining How Transformers Use Context to Build Predictions (2023.acl-long)

Copied to clipboard

Challenge: Existing methods for analyzing input attributions for a model's prediction are unclear how prior words affect the model' s decision throughout the layers.
Approach: They propose a procedure to analyze models for language generation using the Transformer and a comparison of their results with evidence of the linguistic phenomena.
Outcome: The proposed method consistently aligns better than gradient-based and perturbation-based baselines and generates human-like source-target alignments for building predictions.
Analyzing Encoded Concepts in Transformer Language Models (2022.naacl-main)

Copied to clipboard

Challenge: a new framework to analyze how latent concepts are encoded in representations learned in pre-trained lan-guage models is proposed . conceptX uses clustering to discover the encoded concepts and align them with a large set of human-defined concepts.
Approach: They propose a framework to analyze how latent concepts are encoded in representations learned within pre-trained lan-guage models.
Outcome: The proposed framework explains encoded concepts by aligning with human-defined concepts.
Are Transformers a Modern Version of ELIZA? Observations on French Object Verb Agreement (2021.emnlp-main)

Copied to clipboard

Challenge: Recent studies have shown that unsupervised sentence representations of neural networks encode syntactic information by observing that neural language models are able to predict the agreement between a verb and its subject.
Approach: They propose to take an alternative look at these results by studying whether neural networks are able to build an abstract sentence representation rather than capture surface statistical regularities.
Outcome: The proposed model can achieve high accuracy on the long-range French object-verb agreement, indicating a possible flaw in the model's syntactic ability.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations