
Amazon Science
@AmazonScience
Followers
93K
Following
10K
Media
621
Statuses
7K
The latest news and research from Amazon's science community. #AmazonScience
Global
Joined December 2019
Nova Act is now⚡️ enterprise ready ⚡️ and we've added new capabilities to our preview to help you take your prototype to production—with 90%+ reliability across our early enterprise customer use cases!
1
6
31
To improve LLM safety, Amazon researchers demonstrate that using multiple AI agents to generate and refine "chain of thought" training data boosts benchmark performance by 29% on average: #ACL2025 #ACL2025NLP #NLProc.
amazon.science
Using ensembles of agents to generate and refine interactions annotated with chains of thought improves performance on a battery of benchmarks by an average of 29%.
0
0
2
RT @DavidZapolsky: The Presidential AI Challenge will empower America's brightest students, educators & mentors to solve real-world problem….
orise.orau.gov
The Presidential AI Challenge seeks to inspire young people and educators to create AI-based innovative solutions to community challenges while fostering AI interest and competency.
0
2
0
Amazon researchers introduce the cost-to-serve-software (CTS-SW) metric to measure software development efficiency and costs. Their study shows AI-powered developer tools reduce costs and boost productivity across the full lifecycle:
amazon.science
New cost-to-serve-software metric that accounts for the full software development lifecycle helps determine which software development innovations provide quantifiable value.
0
0
3
Join us at the inaugural NY Reinforcement Learning (NYRL) Workshop! Connect with leading #RL researchers from academia & industry on Sept 12 @ Amazon JFK27 NYC. Submit by Aug 21 💡 Learn more:
1
0
1
We're excited to kick off @aclmeeting in Vienna! 🇦🇹. Amazon is showcasing nearly 30 papers advancing computational linguistics and natural language processing. Find us at booth #18 to connect with Amazon researchers and explore our latest work:
1
0
8
RT @akhil_bagaria: 🚨New workshop alert 🚨. Calling all RL researchers in the New York area 🧠🗽. Present your work at the first-ever New York….
0
15
0
RT @RohitPrasadAI: Recently showed how Nova Act is helping Alvee Health streamline the notoriously complex process of applying for public b….
0
2
0
RT @siebelschool: #ILLINOIS PurpCorn-PLAN has won the @amazon Nova AI Challenge, claiming an impressive $250,000 prize! 🔶🏆🔷. This team of 1….
amazon.science
University teams battle to harden and hack AI coding assistants in head-to-head tournament
0
4
0
✍️ New from Amazon's AGI Lab: Why do knowledge workers spend more time managing digital tools than thinking? @drperszyk explores the modern productivity paradox and presents a fresh vision for human-AI collaboration:
labs.amazon.science
In 1946, two young scientists at the University of Pennsylvania introduced ENIAC, the world’s first digital computer. It was the size of a room, consumed the power of a small town, and had one...
1
0
10
RT @DavidZapolsky: Amazon welcomes @WhiteHouse's AI Action Plan, which closely aligns with our efforts and investments. To maintain U.S. AI….
0
23
0
Introducing Mitra: a foundation model from Amazon researchers that outperforms traditional methods for tabular data by learning from diverse synthetic priors. Mitra uses in-context learning to adapt to new tasks without separate models for each dataset:
amazon.science
Generating diverse synthetic prior distributions leads to a tabular foundation model that outperforms task-specific baselines.
0
7
21
Amazon researchers developed a new AI architecture that cuts inference time by 30% by selecting only task-relevant neurons, similar to how the brain uses specialized regions for specific tasks.
amazon.science
Language models inspired by specialized processing regions in the brain offer significant time and cost savings.
2
4
25
New independent research from Aymara tested 20 leading language models against 10 real-world safety risks. The findings? Amazon Nova emerged as a leader, demonstrating strong performance across misinformation, impersonation, and privacy categories.
medium.com
Most Popular AI Models Are Failing Key Real-World Safety Tests. Here’s the Data.
0
0
2
That concludes another successful ICML. Thank you to everyone who engaged with our work and connected with our team throughout the conference. Until next time! #ICML2025
2
0
15