Date: Thursday, April 23

Time: 12:00 PM

Room # 204C

Presenter: Satyaki Chakraborty

Abstract: We are excited to present Amazon Nova, a groundbreaking portfolio of AI offerings that deliver frontier intelligence and industry-leading price performance. Nova is built on advanced AI technologies originally developed for Amazon's internal applications, such as Alexa+, Amazon Ads, and AWS Marketplace, and is now available to AWS customers. Amazon Nova includes Nova models, fast and cost-effective foundation models for text and multimodal needs; Nova Forge, a new service to build your own frontier models; and Nova Act, a new service to build agents that automate browser-based UI workflows powered by a custom Nova 2 Lite model. These models and services have built-in controls for the safe and responsible use of AI, delivering robust protections, content filters, and policy-aligned behaviors to meet compliance requirements. During our demo, a product engineer and researcher will showcase the product and the science behind it. We will demonstrate how Nova models can be customized to deliver responses that reflect industry expertise, powering interactive chat interfaces, Retrieval-Augmented Generation (RAG) systems, agentic applications, video analysis, and UI workflow automation solutions. We will also highlight the multimodal capabilities of Nova, which accept text, image, or video inputs to generate text output, and the creative content generation models that accept text and image inputs to generate image or video output. Amazon Nova has been adopted by tens of thousands of customers across industries, delivering measurable impact with cost savings and gains in productivity, automation, and quality with real-world deployments.

Date: Friday, April 24

Time: 12:00PM

Room # 204C

Presenter: Udita Patel

Abstract: LLM-based agents struggle to execute complex, multi-step Standard Operating Procedures (SOPs) that are fundamental to industrial automation. Existing benchmarks fail to capture the procedural complexity and tool orchestration demands of real-world workflows. We introduce SOP-Bench, a benchmark of 2,000+ tasks from human expert-authored SOPs across 12 business domains (healthcare, logistics, finance, content moderation, etc.). Using a human-AI collaborative framework, experts crafted authentic SOPs while AI generated artifacts (tools, APIs, datasets), all human-validated, yielding realistic tasks with executable interfaces and ground-truth outputs.
SOP-Bench serves as a research enabler for systematically investigating agent architectures, model capabilities, and deployment considerations across diverse procedural tasks. In this talk, we discuss the challenges of real world SOPs and a framework on evaluating them.

This workshop brings together researchers to tackle a timely question: how do we build the algorithmic foundations for powerful and reliable self-improving AI systems? As LLM agents rewrite their own code and prompts, scientific discovery pipelines schedule continual fine-tuning, and robotics stacks patch controllers from streaming telemetry, recursive self-improvement is moving from thought experiments to production. This workshop examines algorithms for self-improvement across experience learning, synthetic data pipelines, multimodal agentic systems, weak-to-strong generalization, and inference-time scaling—focused on loops that actually get better and can show it. Contributions are organized around five lenses: change targets, temporal regime, mechanisms and drivers, operating contexts, and evidence of improvement. The workshop is paradigm-agnostic, welcoming work on foundation models, agent frameworks, robots, learning algorithms, control and program synthesis, and the data, infrastructure, and evaluation tooling that enable recursive self-improvement.

Website: https://iclr.cc/virtual/2026/workshop/10000796

This workshop brings together researchers from machine learning, economics, and computer science to explore the growing synergy between modern AI—particularly foundation models—and the classical fields of Mechanism Design and Strategic Decision Making. It investigates how AI methods can redefine, extend, and automate core problems in these domains, spanning novel applications, theoretical foundations for AI-driven approaches, and real-world case studies from industry. The goal is to foster interdisciplinary collaboration and chart the future of intelligent economic systems.

Website: https://alimama-tech.github.io/aims-2026/

This workshop will explore the challenges and opportunities for improving deduction, induction, and abduction capabilities of LLMs, implementing symbolic representation and reasoning via LLMs, avoiding logical contradictions across responses to multiple related questions, enhancing LLM reasoning by leveraging external logical solvers, and benchmarking LLM logical reasoning and consistencies. As LLMs continue to expand their role in AI research and applications, this workshop will serve as a platform to discuss and refine the methods for advancing logical reasoning within LLMs.

Website: https://sites.google.com/view/iclr-2026-llmreasoning

Website: https://tsalm-workshop.github.io/

The financial domain is undergoing rapid transformation driven by advances in artificial intelligence. Building on last year’s "Advances in Financial AI: Opportunities, Innovations, and Responsible AI" workshop, this second edition will focus particularly on the emergence of agentic systems in finance (autonomous or semi-autonomous agents, decision-making systems, multi-agent interactions), and the imperative of responsibility (ethics, fairness, accountability, transparency, robustness, regulation). This workshop aims to bring together researchers, practitioners, and policymakers to explore both the opportunities and risks of agentic financial AI systems, to share recent innovations, and to work towards foundations and best practices that ensure such systems are safe, trustworthy, and socially aligned.

Website: https://iclr.cc/virtual/2026/workshop/10000788

This workshop builds on a growing community of machine learning, neuroscience, and cognitive science researchers to tackle a new question: what can we do with alignment? This year's workshop pivots from measuring representational alignment to exploring its affordances, with two interdisciplinary focus areas. The first, neural control, examines when alignment allows meaningful intervention on a system's behavior—connecting mechanistic interpretability and model steering in AI with understanding how neural activity gives rise to function in the brain. The second, downstream behavior, explores how alignment enables targeted control over how representations are deployed for specific tasks, moving beyond "does the model know X?" to "can we steer when and how it applies X?" T

Website: https://representational-alignment.github.io/2026/

Amazon speaker: Danielle Perszyk

This workshop addresses methods for ensuring the reliability of machine learning systems under distribution drift. Such drift may occur as temporal, label, concept, domain, or representation shift, each presenting unique challenges for monitoring and adaptation.

Website: https://sites.google.com/view/iclr-2026-workshop-cao/

Website: https://lifelongagent.github.io/