Papers by Alvaro Soto

8 papers
Tracr-Injection: Distilling Algorithms into Pre-trained Language Models (2025.findings-acl)

Copied to clipboard

Challenge: Recent efforts to characterize symbolic abilities of the transformer architecture have shown that the tasks that can be implemented in RASP are uncommon to learn from natural unsupervised data.
Approach: They propose a programming language, called RASP, which can be directly compiled into transformer weights to implement these algorithms.
Outcome: The proposed method improves out-of-distribution performance compared to baselines, indicating that indeed a more symbolic mechanism is taking place in the inner workings of the model.
Augmenting BERT-style Models with Predictive Coding to Improve Discourse-level Representations (2021.emnlp-main)

Copied to clipboard

Challenge: Existing language models do not produce suitable representations at the discourse level.
Approach: They propose to augment BERT-style language models with a mechanism that allows them to learn suitable discourse-level representations by incorporating top-down connections that operate at the intermediate layers of the network.
Outcome: The proposed approach improves in 6 out of 11 tasks by detecting discourse relationship detection.
Translating Navigation Instructions in Natural Language to a High-Level Plan for Behavioral Robot Navigation (D18-1)

Copied to clipboard

Challenge: Existing models for translating free-form natural language instructions to a high-level plan for behavioral robot navigation are difficult due to the variability in the way people describe routes.
Approach: They propose an end-to-end deep learning model for translating free-form natural language instructions to a high-level plan for robot navigation.
Outcome: The proposed model significantly outperforms baseline approaches on a new dataset containing 10,050 pairs of navigation instructions.
A Memory Model for Question Answering from Streaming Data Supported by Rehearsal and Anticipation of Coreference Information (2023.findings-acl)

Copied to clipboard

Challenge: Existing question answering methods assume that the input content can always be accessed while answering the question.
Approach: They propose a model that performs rehearsal and anticipation while processing inputs to memorize important information for question answering tasks from streaming data.
Outcome: The proposed model improves on short-sequence (bAbI) and large-squence textual (NarrativeQA) and video (ActivityNet-QA) question answering datasets.
Evaluation Benchmarks for Spanish Sentence Representations (2022.lrec-1)

Copied to clipboard

Challenge: Existing and newly constructed datasets address different tasks from various domains.
Approach: They propose to use Spanish SentEval and Spanish DiscoEval to evaluate stand-alone and discourse-aware sentence representations.
Outcome: The proposed benchmarks evaluate the capabilities of stand-alone and discourse-aware sentence representations in Spanish and show that they are more robust and comparable than previous benchmarks.
Inspecting the concept knowledge graph encoded by modern language models (2021.findings-acl)

Copied to clipboard

Challenge: Pre-trained language models are used to solve tasks such as summarization and information retrieval.
Approach: They propose to use word embeddings, text generators, context encoders to extract underlying knowledge graphs of nine influential language models.
Outcome: The proposed model is able to encode word embeddings, text generators, and context encoders, but suffers from several inaccuracies.
Extracting and Encoding: Leveraging Large Language Models and Medical Knowledge to Enhance Radiological Text Representation (2024.findings-acl)

Copied to clipboard

Challenge: Advancing representation learning in specialized fields like medicine remains challenging due to the scarcity of expert annotations for text and images.
Approach: They propose a Fact Extractor that leverages large language models to extract factual statements from radiology reports.
Outcome: The proposed framework outperforms current state-of-the-art methods in sentence ranking, natural language inference, and label extraction tasks.
How Relevant is Selective Memory Population in Lifelong Language Learning? (2022.aacl-short)

Copied to clipboard

Challenge: Existing approaches to lifelong language learning rely on sparse experience replay to prevent catastrophic forgetting.
Approach: They propose to use a selective memory population to store a uniform number of samples from the entire data stream to improve model performance.
Outcome: The proposed methods show that they are relevant for lifelong language learning tasks, especially for low memory size, and consistent with computer vision studies.

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