Papers by Richard Johansson

13 papers
Surface-Based Retrieval Reduces Perplexity of Retrieval-Augmented Language Models (2023.acl-short)

Copied to clipboard

Challenge: Augmenting language models with a retrieval mechanism has been shown to improve performance while keeping the number of parameters low.
Approach: They propose to augment language models with a retrieval mechanism by replacing semantic retrieval with BM25 . they find that the model's performance is better explained by surface-level similarities, they say .
Outcome: The proposed method reduces perplexity and lowers the number of parameters while keeping the number low.
Transformer-based Swedish Semantic Role Labeling through Transfer Learning (2024.lrec-main)

Copied to clipboard

Challenge: Semantic Role Labeling (SRL) is a task in natural language understanding where the goal is to extract semantic roles for a given sentence.
Approach: They propose to build a Transformer-based SRL system for Swedish by exploring multilingual and cross-lingual transfer learning methods and leveraging the Swedish FrameNet resource.
Outcome: The proposed model outperforms two different cross-lingual transfer models and shows that the multilingual learning outperformed the other models.
Training a Swedish Constituency Parser on Six Incompatible Treebanks (2020.lrec-1)

Copied to clipboard

Challenge: Syntactic parsing is a widely used intermediate step in several natural language processing tasks.
Approach: They propose to use a function-tagged constituent treebank for Swedish which includes discontinuous constituents to improve the accuracy.
Outcome: The proposed parser can be trained on additional treebanks that use other annotation models.
Benchmarking Debiasing Methods for LLM-based Parameter Estimates (2025.emnlp-main)

Copied to clipboard

Challenge: Large language models (LLMs) are expensive yet powerful ways to annotate text, and can be inconsistent when compared with experts.
Approach: They propose to combine LLM annotations with a limited number of expensive expert annotations to produce valid estimates.
Outcome: The proposed methods produce consistent estimates under theoretical assumptions, but they are not comparable across finite datasets.
What do Models Learn From Training on More Than Text? Measuring Visual Commonsense Knowledge (2022.acl-srw)

Copied to clipboard

Challenge: Existing evaluation methods to measure what language models learn from multimodal training are lacking.
Approach: They propose two evaluation tasks to measure commonsense knowledge in language models by using visual data to evaluate multimodal models and unimodal baselines.
Outcome: The proposed evaluation tasks show that training on a visual modality improves on the visual commonsense knowledge in language models.
What Happens to a Dataset Transformed by a Projection-based Concept Removal Method? (2024.lrec-main)

Copied to clipboard

Challenge: A number of recent methods have used linear projections to remove information about a concept from a language representation.
Approach: They propose to use linear projections to remove a concept from a language representation to create a transformed set of word embeddings.
Outcome: The proposed methods inject strong statistical dependencies into the transformed datasets.
The Effect of Scaling, Retrieval Augmentation and Form on the Factual Consistency of Language Models (2023.emnlp-main)

Copied to clipboard

Challenge: Large Language Models (LLMs) are useful interfaces to factual knowledge, but their usefulness is limited by their tendency to deliver inconsistent answers to semantically equivalent questions.
Approach: They evaluate the effectiveness of up-scaling and augmenting the LM with a passage retrieval database to reduce inconsistency.
Outcome: The proposed models reduce inconsistency but retrieval augmentation is more efficient.
On the Generalization Ability of Retrieval-Enhanced Transformers (2023.findings-eacl)

Copied to clipboard

Challenge: Recent work on retrieval-augmented language models has shown impressive results . performance gains from retrieval to a large extent originate from overlapping tokens between the database and test data, suggesting less of non-trivial generalization than previously assumed.
Approach: They propose to off-load memory from trainable weights to a retrieval database and compare it to larger models with a larger model.
Outcome: The proposed model outperforms GPT-3 and Jurassic-1 on the Pile at 4% of the model parameters.
Conceptualizing Treatment Leakage in Text-based Causal Inference (2022.naacl-main)

Copied to clipboard

Challenge: Existing methods to control for text-based confounders rely on assumption that there is no treatment leakage . prior literature has assumed that documents only contain information about confounder, but not about treatment assignment.
Approach: They define the treatment leakage problem and propose methods to mitigate it . they remove treatment-related signal from text in a pre-processing step .
Outcome: The proposed method can mitigate the problem of treatment leakage by removing the treatment-related signal from the text.
Deciphering the Interplay of Parametric and Non-parametric Memory in Retrieval-augmented Language Models (2024.emnlp-main)

Copied to clipboard

Challenge: Generative language models struggle with specialized knowledge that is discussed less frequently on the web.
Approach: They propose to use a model which decides between parametric and non-parametric knowledge to investigate how it uses the information from the context.
Outcome: The proposed model can choose between parametric and non-parametric information, but relies more on context than parametric knowledge.
CUB: Benchmarking Context Utilisation Techniques for Language Models (2026.acl-long)

Copied to clipboard

Challenge: Existing language models (LMs) can be distracted by irrelevant contexts or ignore relevant information that contradicts outdated parametric memory.
Approach: They develop a benchmark to help diagnose CMTs under diverse noisy context conditions within retrieval-augmented generation (RAG) they find that most existing CMT struggle to handle the full spectrum of context types encountered in real-world RAG scenarios.
Outcome: The proposed benchmark compares seven state-of-the-art methods across three datasets and tasks, and shows that many lack the robustness needed to handle the full spectrum of context types encountered in real-world RAG scenarios.
Fact Recall, Heuristics or Pure Guesswork? Precise Interpretations of Language Models for Fact Completion (2025.findings-acl)

Copied to clipboard

Challenge: Language models (LMs) can make a correct prediction based on many possible signals in a prompt, but not all corresponding to recall of factual associations.
Approach: They propose a model-specific recipe for constructing datasets with examples of four different prediction scenarios: generic language modeling, guesswork, heuristics recall and exact fact recall.
Outcome: The proposed model-specific recipe yields distinct results for each scenario.
How to Adapt Pre-trained Vision-and-Language Models to a Text-only Input? (2022.coling-1)

Copied to clipboard

Challenge: Current language models have been criticised for learning language from text alone without connection between words and their meaning.
Approach: They propose to train models on more sources than text to provide the lacking connection between words and their meanings.
Outcome: The proposed model adaptation methods perform differently for different models and unimodal model counterparts perform on par with the VL models regardless of adaptation.

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