Summary of Lofti: Localization and Factuality Transfer to Indian Locales, by Sona Elza Simon (1) et al.
LoFTI: Localization and Factuality Transfer to Indian Locales
by Sona Elza Simon, Soumen Kumar Mondal, Abhishek Singhania, Sayambhu Sen, Preethi Jyothi
First submitted to arxiv on: 16 Jul 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Large language models (LLMs) have been trained on vast amounts of web-scale data, but these datasets often exhibit geographical biases towards English-speaking Western countries. This bias can lead to LLMs producing inaccurate or hallucinated responses when answering queries about other regions. To address this issue, we introduce LoFTI (Localization and Factuality Transfer to Indian Locales), a new benchmark for evaluating an LLM’s localization and factual text transfer capabilities. LoFTI consists of factual statements about entities in source locations worldwide and target locations within India with varying degrees of hyperlocality (country, state, city). We use LoFTI to evaluate Mixtral, GPT-4, and two other Mixtral-based approaches designed for localized factual transfer. Our results show that LoFTI is a high-quality evaluation benchmark, and all models, including GPT-4, produce skewed results across different levels of hyperlocality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models have been trained on huge amounts of data from the internet, but this training data often comes from English-speaking countries. This can cause problems when the model is asked to answer questions about other parts of the world. We created a new way to test how well these models do in answering localized questions. Our test, called LoFTI, includes statements about things like people and places in different parts of India. We used LoFTI to see how well some popular language models did at answering these kinds of questions. The results showed that our test is a good way to evaluate how well language models do this kind of task. |
Keywords
» Artificial intelligence » Gpt
