December 2025

Our latest newsletter features our top blog posts and publications of 2025, deep dives into LLM applications, AI-native 6G networks, and dialogue enhancement technology, along with Amazon Nova announcements at AWS re:Invent.

Best of 2025

Deep dives

Using LLMs to improve Amazon product listings: Through prompt tuning, LLMs adapt to catalogue structures and vocabulary to correct attributes at scale. The system recognizes standard values, collects alternative representations, and detects erroneous entries.

AI-native 6G: From networks to intelligence fabrics: AWS researchers explain how network language models will enable hyper-composed networks that dynamically orchestrate computation, storage, networking, data, and AI resources across providers and jurisdictions.

Dialogue Boost: How Amazon is using AI to enhance TV and movie dialogue: The technology separates speech from background sounds using sub-band neural networks compressed to <1% original size. Results show 86% listener preference and 100% approval from users with hearing loss.

AutoGluon assistant: Zero-code AutoML through multiagent collaboration: MLZero is a multiagent architecture that separates data perception, tool knowledge, execution history, and code generation to enable machine learning automation that works with messy, real-world inputs.

News and updates

Amazon Nova Forge: "Open training” paradigm that empowers everyone to build their own frontier AI: By mixing proprietary data with the data used to train Amazon Nova at each major stage of model development—pretraining, mid-training, and post-training—customers achieve deep domain understanding while preventing "catastrophic forgetting."

Building AI agents that actually work: Introducing Nova Act as a service: Nova Act is now available as an AWS service for building and managing fleets of UI-automation agents at scale. With customers achieving over 90% reliability on UI-based workflows, Nova Act is the fastest and easiest path to production for developers.

Featured publications

• Amazon Nova 2: Multimodal reasoning and generation models
• Multimodal music tokenization with residual quantization for generative retrieval
• CausalFairnessInAction: An open source python library for causal fairness analysis

LinkedIn | X/Twitter | Facebook | Instagram | GitHub | RSS

© 1996-2025 Amazon.com, Inc. or its affiliates | Privacy | Conditions of Use