Reasoning models can generate seven to 10 times as many tokens as necessary on simple tasks, creating unsustainable costs at scale. Amazon's vision for metacognitive AI could fundamentally shift how models allocate computational resources.
Amazon's competitive-agent architecture creates a continuous improvement cycle that develops security protections at machine speed, reducing what typically takes weeks down to hours.
A new approach to reducing carbon emissions reveals previously hidden emission “hotspots” within value chains, helping organizations make more detailed and dynamic decisions about their future carbon footprints.
How agentic systems work under the hood — and how AWS’s new AgentCore framework implements their essential components.
Customer-obsessed science
Research areas
-
November 20, 20254 min readA new evaluation pipeline called FiSCo uncovers hidden biases and offers an assessment framework that evolves alongside language models. -
October 20, 20254 min read -
October 14, 20257 min read
-
October 2, 20253 min read
-
Featured news
Initiative will fund over 100 doctoral students researching machine learning, computer vision, and natural-language processing at nine universities.
The collaboration will advance research in generative AI, robotics, natural language processing and cloud computing while fostering innovation in foundational and emerging technologies.
University teams battle to harden and hack AI coding assistants in head-to-head tournament
David Chang/Getty Images/iStockphoto
Led by David Luan and Pieter Abbeel, the lab will focus on developing new foundational capabilities for enabling useful AI agents.
The company's new state-of-the-art foundation models deliver frontier intelligence and industry-leading price performance.
-
Code@MIT 20252025In A/B testing, statistical power depends on both the variance of estimated impacts and the distribution of true impacts. A low variance metric can have low power if true impacts on the metric tend to be small, while a high variance metric can have high power if true impacts on the metric tend to be large. Traditional power calculations, however, focus solely on the variance of estimated impacts. They compute
-
NeurIPS 2025 Workshop on Bridging Language, Agent, and World Models (LAW)2025We present a framework for uncovering and exploiting dependencies among tools and documents to enhance exemplar artifact generation. Our method begins by constructing a tool knowledge graph from tool schemas—including descriptions, arguments, and output payloads—using a DeepResearch-inspired analysis. In parallel, we derive a complementary knowledge graph from internal documents and SOPs, which is then
-
QCE 20252025The rapid evolution of quantum hardware necessitates an adaptable static analysis framework for validating quantum programs. In this work, we introduce SHARP, a rule-based static analysis framework designed for OpenQASM that decouples hardware-specific constraints from the validation engine. By employing a rule-based approach, SHARP allows quantum computing services to validate programs against evolving
-
Code@MIT 20252025User-randomized A/B testing, while the gold standard for online experimentation, faces significant limitations when legal, ethical, or practical considerations prevent its use. Item-level randomization offers an alternative but typically suffers from high variance and low statistical power due to skewed distributions and limited sample sizes. We here introduce Regular Balanced Switchback Designs (RBSDs)
-
Code@MIT 20252025This paper examines the effectiveness of stratification in experimental design using evidence from multiple large-scale experiments. We analyze data from experiments ranging from approximately 30,000 to 180,000 units across different business contexts. Our results show that pre-stratification and post-stratification achieve virtually identical precision improvements - largest in smaller samples (10% improvement
Collaborations
View allWhether you're a faculty member or student, there are number of ways you can engage with Amazon.
View all
Gretchen Ertl
The program offers unrestricted funds and other resources to support research at academic institutions and non-profit organizations in areas that align with our mission.
A global university competition to drive secure innovation in generative AI technology, which focuses on responsible AI and large language model coding security.
Credit: Wolfram Scheible
We partner with particular academic organizations across the world for deep and sustained collaborations in multiple research areas of mutual interest.
Courtesy of Pai-Ling Yin
We hire world-class academics to work on large-scale technical challenges, while they continue to teach and conduct research at their universities.
