# When Does Aggregating Multiple Skills with Multi-Task Learning Work? A Case Study in Financial NLP

Source: [When Does Aggregating Multiple Skills with Multi-Task Learning Work? A Case Study in Financial NLP](<https://www.sota2.com/research/paper/68fa6f691e209131cea4569f>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-04T04:33:19.130Z

Authors: Jingwei Ni, Zhijing Jin, Qian Wang, Mrinmaya Sachan, Markus Leippold

Published: 2023-05-23T12:37:14.000Z

## Abstract

Multi-task learning (MTL) aims at achieving a better model by leveraging data and knowledge from multiple tasks. However, MTL does not always work -- sometimes negative transfer occurs between tasks, especially when aggregating loosely related skills, leaving it an open question when MTL works. Previous studies show that MTL performance can be improved by algorithmic tricks. However, what tasks and skills should be included is less well explored. In this work, we conduct a case study in Financial NLP where multiple datasets exist for skills relevant to the domain, such as numeric reasoning and sentiment analysis. Due to the task difficulty and data scarcity in the Financial NLP domain, we explore when aggregating such diverse skills from multiple datasets with MTL can work. Our findings suggest that the key to MTL success lies in skill diversity, relatedness between tasks, and choice of aggregation size and shared capacity. Specifically, MTL works well when tasks are diverse but related, and when the size of the task aggregation and the shared capacity of the model are balanced to avoid overwhelming certain tasks.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Sentiment Classification on Malo SC 2013](<https://www.sota2.com/research/sota/sentiment-classification-on-malo-sc-2013.md>): Sentiment Classification; Malo SC 2013; Accuracy: 86.61.

- [Target-based sentiment analysis on TSA 2017](<https://www.sota2.com/research/sota/target-based-sentiment-analysis-on-tsa-2017.md>): Target-Based Sentiment Analysis; TSA 2017; RMSE: 0.2054.

- [Financial Natural Language Processing on FinDATA](<https://www.sota2.com/research/sota/financial-natural-language-processing-on-findata.md>): Financial Natural Language Processing; FinDATA; TSA: 0.2159.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2305.14007>)

[PDF](<https://arxiv.org/pdf/2305.14007.pdf>)

[Implementation](<https://github.com/EdisonNi-hku/MTL4Finance>)