Making fairness in large language models observable, quantifiable, and governable: Amazon researchers developed FiSCo (fairness in semantic context), a three-stage evaluation pipeline that detects whether language models respond fairly to different groups of people defined by sensitive attributes such as gender, race, and age when multiple valid responses to their questions exist: https://amzn.to/4oTahxj
Amazon Science
Research Services
Seattle, Washington 383,488 followers
The latest news and research from Amazon’s science community. #AmazonScience
About us
Amazon Science gives you insight into the company’s approach to customer-obsessed scientific innovation. Amazon fundamentally believes that scientific innovation is essential to being the most customer-centric company in the world. It’s the company’s ability to have an impact at scale that allows us to attract some of the brightest minds in artificial intelligence and related fields. Our scientists continue to publish, teach, and engage with the academic community, in addition to utilizing our working backwards method to enrich the way we live and work. Follow us on LinkedIn and visit our website to get a deep dive on innovation at Amazon, and explore the many ways you can engage with our scientific community. #AmazonScience
- Website
-
https://www.amazon.science
External link for Amazon Science
- Industry
- Research Services
- Company size
- 10,001+ employees
- Headquarters
- Seattle, Washington
- Founded
- 2020
- Specialties
- Artificial Intelligence, Machine Learning, Computer Vision, Cloud, Economics, Sustainability, AI, ML, Conversational AI, Natural Language Processing, NLP, Robotics, Security, Privacy, Information, Knowledge Management, Operations, Scientific Research, Search, Amazon, and Alexa
Updates
-
Pretrained time series models have enabled inference-only forecasting systems that produce accurate predictions without task-specific training. However, existing approaches largely focus on univariate forecasting, limiting their applicability in real-world scenarios where multivariate data and covariates play a crucial role. We present Chronos-2, a pretrained model capable of handling univariate, multivariate, and covariate-informed forecasting tasks in a zero-shot manner. Chronos-2 employs a group attention mechanism that facilitates in-context learning (ICL) through efficient information sharing across multiple time series within a group, which may represent sets of related series, variates of a multivariate series, or targets and covariates in a forecasting task. These general capabilities are achieved through training on synthetic datasets that impose diverse multivariate structures on univariate series. Chronos-2 delivers state-of-the-art performance across three comprehensive benchmarks: fev-bench, GIFT-Eval, and Chronos Benchmark II. On fev-bench, which emphasizes multivariate and covariate-informed forecasting, Chronos-2’s universal ICL capabilities lead to substantial improvements over existing models. On tasks involving covariates, it consistently outperforms baselines by a wide margin. Case studies in the energy and retail domains further highlight its practical advantages. The in-context learning capabilities of Chronos-2 establish it as a general-purpose forecasting model that can be used “as is” in real-world forecasting pipelines. Chronos-2 model card: https://amzn.to/3LwRwkp Deploy Chronos-2 on Amazon SageMaker: https://amzn.to/3LzIyD1 Chronos-2 technical report: https://amzn.to/3JrDsIs Chronos GitHub Repository: https://amzn.to/3Vop4E5
-
Amazon Science reposted this
Curious how AI agents really work? 🤖 🔍 💡 https://go.aws/43rhxaV Our VP/Distinguished Engineer Marc Brooker breaks down the technology & shows how Amazon Bedrock AgentCore can help move your agents from prototype to production faster. This comprehensive agentic platform helps you build, deploy, & operate highly capable agents at scale, without infrastructure management. #AmazonBedrock
-
At this year's Amazon Machine Learning Conference (AMLC), Amazon researchers gathered to showcase high-quality science, connect with colleagues with similar research interests, and learn innovative techniques. Since 2013, the annual internal conference has provided learning and teaching opportunities for scientists across the company.
-
-
Announcing a private AI bug bounty program to strengthen the Amazon Nova foundation models. Building on the success of Amazon's public bug bounty program, which has surfaced over 30 validated findings and awarded over $55,000 in rewards, the new program will partner with the broader security and academic research communities to strengthen the security of select Amazon AI models and applications including Nova, with qualified participants earning monetary rewards ranging from $200 to $25,000.
-
Amazon researchers reveal previously hidden emission “hotspots” within supply chains using a new mathematical approach. Their freely available dataset shows emissions at each tier of the value chain, enabling organizations to create dynamic, scenario-based models for strategic supplier and technology decisions instead of relying on a static snapshot: https://amzn.to/47uPXMz
-
-
Amazon Science reposted this
Amazon is investing in the future of AI development by supporting more than 100 doctoral students across nine top universities, through our $68 million AI PhD Fellowship program. This initiative represents our commitment to advancing the frontiers of artificial intelligence through machine learning, computer vision, and natural language processing research. Each university will receive $1.1 million annually to support tuition, stipends, and research expenses, ensuring these brilliant minds can focus on breakthrough discoveries. During my time as an affiliate instructor and as Director of the University of Washington Foundation, I’ve witnessed firsthand the transformative power of academic-industry partnerships and am excited about this program's potential to accelerate AI innovation while nurturing the next generation of researchers. The University of Washington is among the participating institutions, and this partnership exemplifies how we can bridge the gap between cutting-edge research and real-world applications. Read more about how this fellowship program embodies our commitment to fostering innovation that benefits society while maintaining the highest standards of safety and responsibility: https://lnkd.in/gb5mPkwN
-
Last week wrapped up this year's International Conference on Computer Vision in Hawaii. You can learn more about our computer vision research here: https://lnkd.in/eAXrWDRs Thank you to everyone who stopped by! #ICCV2025
-
-
Amazon is launching the next chapter of the Amazon Nova AI Challenge — an annual university competition advancing the science of responsible, real-world AI. The 2026 Challenge: Trusted Software Agents asks student teams to build and evaluate AI agents that can plan, build, and test software safely and reliably. As generative AI expands from code generation to complex application development, this challenge focuses on improving both utility and trustworthiness in tandem. Ten university teams will compete to design systems that reflect real engineering workflows, demonstrating measurable progress in both performance and safety. Applications open November 10, 2025. Learn more: http://amzn.to/43Ef8cZ