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.

Similar Papers

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.
Additive Compositionality of Word Vectors (D19-55)

Copied to clipboard

Challenge: Existing research on justifying additive compositionality of word embedding models requires a rather strong assumption of uniform word distribution.
Approach: They propose to relax the assumption of uniform word distribution and propose more realistic conditions for proving additive compositionality.
Outcome: The proposed model improves on word similarity and noisy sentence similarity.
Simple and effective data augmentation for compositional generalization (2024.naacl-long)

Copied to clipboard

Challenge: Compositional generalization is the ability of a system to correctly predict the meaning of complex sentences when trained on simpler sentences.
Approach: They propose to use data augmentation methods to generate additional training data by sampling from an augmentation distribution to generalize to the out-of-distribution test data.
Outcome: The proposed method outperforms existing methods that sampled from the training distribution and outperformed existing methods.
What’s in Your Embedding, And How It Predicts Task Performance (C18-1)

Copied to clipboard

Challenge: Attempts to find a single technique for general-purpose intrinsic evaluation of word embeddings have so far not been successful.
Approach: They propose a method that quantifies interpretable characteristics of word vector neighborhoods and shows how they correlate with performance on 14 extrinsic and intrinsic task datasets.
Outcome: The proposed approach enables multi-faceted evaluation, parameter search, and generally – a more principled, hypothesis-driven approach to development of distributional semantic representations.
Montague semantics and modifier consistency measurement in neural language models (2025.coling-main)

Copied to clipboard

Challenge: Existing studies on distributional language models have been focused on linguistics and their relationship with semantic formalisms for decades.
Approach: They propose a method for measuring compositional behavior in contemporary language embedding models by introducing three new tests inspired by Montague semantics.
Outcome: The proposed method measures compositional behavior in language embedding models on adjectival modifier phenomena in adjective-noun phrases.
Combine to Describe: Evaluating Compositional Generalization in Image Captioning (2022.acl-srw)

Copied to clipboard

Challenge: Recent work on compositionality has focused on the ability to combine simpler concepts to understand & generate arbitrarily more complex conceptual structures.
Approach: They propose to use a set of image captioning models to benchmark their compositional generalization properties.
Outcome: The proposed models do not generalize in terms of systematicity and productivity, but are robust to synonym substitutions.
On Evaluating Multilingual Compositional Generalization with Translated Datasets (2023.acl-long)

Copied to clipboard

Challenge: a growing amount of research investigating compositional generalization in NLP is done on English . a critical semantic distortion is a limitation of the translation of datasets .
Approach: They propose to translate a dataset for evaluating compositional generalization in semantic parsing.
Outcome: The proposed benchmarks show that the translation of the MCWQ dataset suffers from semantic distortion.
Building Static Embeddings from Contextual Ones: Is It Useful for Building Distributional Thesauri? (2022.lrec-1)

Copied to clipboard

Challenge: contextual language models are dominant in the field of Natural Language Processing, but they are not suitable for all uses.
Approach: They propose a method for building word or type-level embeddings from contextual models . they evaluate a large set of English nouns from the perspective of extracting semantic similarity relations .
Outcome: The proposed method can be used to build word or type embeddings from contextual models . it can be exploited for a wide set of English nouns, showing it can improve distributional thesauri .
Long Story Short: Disentangling Compositionality and Long-Caption Understanding in Contrastive VLMs (2026.findings-acl)

Copied to clipboard

Challenge: Existing benchmarks for vision-language models treat compositionality and long-caption understanding in isolation.
Approach: They analyze when compositional reasoning and long-caption understanding transfer across tasks and when this relationship fails.
Outcome: The proposed model can generalize on poorly grounded captions and with strong visual grounding, while architectural choices can limit compositional learning.
Understanding the Influence of Synthetic Data for Text Embedders (2025.findings-acl)

Copied to clipboard

Challenge: Recent advances in general purpose text embedders have been driven by training on synthetic training data.
Approach: They propose to use GPT-4 to produce high quality synthetic data that expands existing training datasets for embeddings to new tasks.
Outcome: The proposed dataset is high quality and leads to consistent improvements in performance.

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