Reactor and the Amazon Neuron Science team developed a kernel-centric optimization strategy that enables real-time, low-latency video generation on AWS Trainium chips, according to Reactor co-founder …
Amazon and Stanford University launched the Stanford and Amazon Research Initiative, a formal framework for joint research in AI, energy, and healthcare, announced at an event on Stanford's campus on …
Amazon Bio Discovery published three papers on AI-driven antibody design, including a new sequence-based binding affinity predictor called MochiBind that ranks candidate antibodies by relative binding…
A new paper, "What fits (into few tokens) doesn't overfit: Compression and generalization in ML research agents," argues that LLM-based research agents avoid overfitting on heavily reused benchmarks b…
A new method from Amazon scientists, presented at the International Conference on Machine Learning (ICML), improves LLM-as-a-judge evaluation by accounting for correlations between judges' outputs, ou…
Amazon scientists released SOP-Bench, the first benchmark pairing genuine enterprise standard operating procedures with functioning tools and ground-truth answers to measure AI agents' ability to exec…
Amazon's Automated Reasoning Group (ARG), launched in 2016, has grown from a small research team into a production service processing billions of queries daily, using mathematical logic to prove the c…
AWS has launched the AWS Trainium Frontier competition, inviting academic and industry labs to co-design language model architectures and custom kernels on its purpose-built Trainium AI chips. Partici…
Amazon Web Services (AWS) announced 34 new recipients of its Build on Trainium research awards, a $110 million credit program supporting AI research on its Trainium chips. The Fall 2025 cycle focused …
Amazon researchers presented ControlG, a framework for multi-objective graph self-supervised learning, at ICML 2026. ControlG uses a proportional-integral-derivative (PID) controller from industrial c…
OpenAI released PatientAgentBench, a clinician-vetted benchmark for evaluating patient-facing AI agents across six safety and workflow dimensions, finding that even capable foundation models fail to m…
Amazon is making the largest single donation in the history of the Lean Focused Research Organization (FRO) to support the development of Lean, a programming language that enables mathematical proofs …
OpenAI released Turnstile, a Rust-based proxy that records exact token-level histories of agentic interactions to improve reinforcement learning for language models. The tool prevents retokenization d…
Amazon is investing in next-generation nuclear technology using TRISO fuel particles to meet the rising energy demands of AI infrastructure and cloud computing. TRISO particles, which are tiny, robust…
Amazon Web Services has launched the M9g and M9gd EC2 instances powered by Graviton5 CPUs, which double core counts to 192 and introduce the Nitro Isolation Engine — the first formally verified hyperv…
Amazon announced the general availability of its new M9g and M9gd EC2 instances, the first products powered by the Graviton5 processor, which delivers up to 25% better computational performance and up…
Amazon's 2026 Project Eluna deploys agentic AI models in fulfillment centers to autonomously plan, use tools, and execute multistep operational tasks, moving AI beyond digital applications into physic…
Amazon researchers have identified a fundamental performance bottleneck in AI agent systems, formalizing it as the "intent-execution gap" — the mismatch between what an AI model intends to do and what…
Amazon's AGI group found that human experts achieved only 60.8% accuracy when verifying claims from AI-generated research reports, revealing that static ground truth datasets are insufficient for eval…
Amazon announced 70 recipients of its Amazon Research Awards from 49 universities across 11 countries, funding research in AI for information security, agentic AI, automated reasoning, cryptography, c…