Amazon Science

The Amazon Science website showcases how the company focuses on scientific innovation with a strong emphasis on customer satisfaction. Amazon is committed to being the most customer-focused company globally, and it sees scientific innovation as a key part of that mission. By making a significant impact on a large scale, Amazon draws in top talent from fields like artificial intelligence and machine learning.

International

English

Blog, Online/Digital

Outlet metrics

Domain Authority

Ranking

Global

#173605

United States

#84155

Science and Education/Math

#158

Traffic sources

Monthly visitors

Articles

Do LLMs recognize your preferences? Evaluating personalized preference following in LLMs

1 month ago | amazon.science | Siyan Zhao |Mingyi Hong |Yang Liu |Devamanyu Hazarika

Large Language Models (LLMs) are increasingly used as chatbots, yet their ability to personalize responses to user preferences remains limited. We introduce PREFEVAL, a benchmark for evaluating LLMs’ ability to infer, memorize and adhere to user preferences in a long-context conversational setting. PREFEVAL comprises 3,000 manually curated user preference and query pairs spanning 20 topics.
Uncertainty-aware fusion: An ensemble framework for mitigating hallucinations in large language models

1 month ago | amazon.science | Prasenjit Dey |Srujana Merugu |Varun Kumar

Large Language Models (LLMs) are known to hallucinate and generate non-factual outputs which can undermine user trust. Traditional methods to directly mitigate hallucinations, such as representation editing and contrastive decoding, often require additional training data and involve high implementation complexity. While ensemble-based approaches harness multiple LLMs to tap into the "wisdom of crowds", these methods overlook uncertainties in individual model responses.
Using instruction-tuned LMs for scalable use case-based shopping - Where customers meet their needs

Jan 14, 2025 | amazon.science | Rajdeep Mukherjee |Sonali Singh |Sachin Farfade |Xinhua Ling

Products on e-commerce platforms are usually organized based on seller-provided product attributes. Customers looking for a product typically have certain needs or use cases in mind, such as headphones for gym classes, or a printer for school projects. However, they often struggle to map these use cases to product attributes, thereby failing to find the product they need.
Evaluating auto-complete ranking for diversity and relevance

Jan 14, 2025 | amazon.science | Sonali Singh |Sachin Farfade |Prakash Mandayam Comar |Xinhua Ling

Traditional Query Auto-completion (QAC) systems optimise for query relevance based on past user interactions. This approach excels at surfacing frequently searched queries, but ensuring a diverse range of suggestions and incorporating new products or trends often requires post-processing heuristics. This limitation stems from relying on user search logs, which may not fully capture the evolving product landscape.
Zero-resource speech translation and recognition with LLMs

Dec 24, 2024 | amazon.science | Karel Mundnich |Xing Niu |Prashant Mathur |Srikanth Ronanki

Despite recent advancements in speech processing, zero-resource speech translation (ST) and automatic speech recognition (ASR) remain challenging problems. In this work, we propose to leverage a multilingual Large Language Model (LLM) to perform ST and ASR in languages for which the model has never seen paired audio-text data.
Neural flows: Efficient alternative to neural ODEs

Nov 21, 2024 | amazon.science | Joanna Sommer |Syama Sundar Rangapuram |Tim Januschowski |Marin Bilos

Marin Bilos, Joanna Sommer, Syama Rangapuram, Tim Januschowski, Stephan Günnemann Copy BibTeX Neural ordinary differential equations describe how values change in time. This is the reason why they gained importance in modeling sequential data, especially when the observations are made at irregular intervals. In this paper we propose an alternative by directly modeling the solution curves — the flow of an ODE — with a neural network.
Neural flows: Efficient alternative to neural ODEs

Nov 21, 2024 | amazon.science | Joanna Sommer |Syama Sundar Rangapuram |Tim Januschowski |Marin Bilos

Marin Bilos, Joanna Sommer, Syama Rangapuram, Tim Januschowski, Stephan Günnemann Copy BibTeX Neural ordinary differential equations describe how values change in time. This is the reason why they gained importance in modeling sequential data, especially when the observations are made at irregular intervals. In this paper we propose an alternative by directly modeling the solution curves — the flow of an ODE — with a neural network.
RefChecker: Reference-based fine-grained hallucination checker and benchmark for large language models

Oct 29, 2024 | amazon.science | Xiangkun Hu |Dongyu Ru |Tianhang Zhang |Zheng Zhang

Large Language Models (LLMs) have shown impressive capabilities but also a concerning tendency to hallucinate. This paper presents REFCHECKER, a framework that introduces claim-triplets to represent claims in LLM responses, aiming to detect fine-grained hallucinations. In REFCHECKER, an extractor generates claim-triplets from a response, which are then evaluated by a checker against a reference.
Unraveling the gradient descent dynamics of transformers

Oct 21, 2024 | amazon.science | Boran Han |Shuai Zhang |Jie Ding |Bingqing Song

While the Transformer architecture has achieved remarkable success across various domains, a thorough theoretical foundation explaining its optimization dynamics is yet to be fully developed. In this study, we aim to bridge this understanding gap by answering the following two core questions: (1) Which types of Transformer architectures allow Gradient Descent (GD) to achieve guaranteed convergence?
Training LLMs to better self-debug and explain code

Oct 21, 2024 | amazon.science | Nan Jiang |Xiaopeng Li |Shiqi Wang |Qiang Zhou

In the domain of code generation, self-debugging is crucial. It allows LLMs to refine their generated code based on execution feedback. This is particularly important because generating correct solutions in one attempt proves challenging for complex tasks.