Agents in the Wild: Safety and Security

More workshops at International Conference on Learning Representations 2026

• Algorithmic Fairness Across Alignment Procedures and Agentic Systems — Examines fairness and bias in the context of RLHF, constitutional AI, and agentic systems -...
• Algorithmic Fairness in AI Systems — Examines fairness, bias, and equity across AI alignment techniques and autonomous agent systems.
• AI for Accelerated Materials Design — Covers machine learning methods for materials discovery, property prediction, and inverse design - from graph...
• AI for Materials Design — Applies machine learning to accelerate discovery of new materials for energy, medicine, and advanced manufacturing....
• AI for Peace — Explores how AI can support conflict prevention, humanitarian response, and peacebuilding applications.
• AI for Partial Differential Equations — Applies machine learning to solving and simulating partial differential equations in physics, climate, and engineering....
• Navigating and Addressing Data Problems for Foundation Models — Tackles the data side of foundation model development: curation, deduplication, quality filtering, data contamination detection,...
• Data Problems for Foundation Models — Tackles data quality, curation, scaling, and contamination challenges in training and evaluating foundation models.
• Foundation Models for Science: Real-World Impact and Science-First Design — Brings together researchers applying large foundation models to scientific discovery, with a focus on real-world...
• Foundation Models for Science — Examines how foundation models can be designed and evaluated for scientific discovery across physics, chemistry,...
• Integrating Generative and Experimental Platforms for Biomolecular Design — Bridges wet lab biology and generative AI, focusing on how active learning loops between computational...
• Generative AI in Genomics — Applies generative models to genomic sequence design, protein structure prediction, and biological data synthesis.
• Geometry-Grounded Representation Learning — Develops representations and generative models grounded in geometric structure for 3D, spatial, and equivariant learning....
• Geometry-grounded Representation Learning and Generative Modeling — Explores how geometric inductive biases - symmetry, equivariance, manifold structure - can improve representation learning...
• From Human Cognition to AI Reasoning — Draws on cognitive science and psychology to inform AI reasoning research, exploring what human mental...
• Lifelong Agents: Learning, Aligning, Evolving — Investigates how AI agents can continuously learn new tasks without forgetting old ones, stay aligned...
• Lifelong Agents Workshop — Addresses continual learning, alignment, and adaptation in AI agents that operate and improve over extended...
• Learning Meaningful Representations of Life — Focuses on representation learning for biological data - from protein structures to genomics to single-cell...
• Meaningful Representations of Life — Applies representation learning to biological data including proteins, genomics, and single-cell sequencing.
• Memory for LLM-Based Agentic Systems — Addresses how LLM-based agents can store, retrieve, and reason over information across long horizons -...
• Memory for LLM-Based Agentic Systems — Studies external memory, retrieval augmentation, and long-context mechanisms for LLM-based agents.
• Machine Learning for Remote Sensing — Covers deep learning for satellite imagery, aerial data, and geospatial analysis - including foundation models...
• Machine Learning for Remote Sensing — Advances ML methods for satellite imagery analysis including land cover mapping, change detection, and weather...
• Machine Learning for Genomics Explorations — Focuses on applying modern deep learning to genomic data: sequence models for DNA and RNA,...
• Machine Learning for Genomics — Advances ML methods for genomic sequence analysis, gene expression, variant calling, and multi-omics integration.
• Multimodal Intelligence: Next Token Prediction and Beyond — Explores architectures and training paradigms for multimodal models that go beyond next-token prediction, including diffusion-based...
• Multimodal Intelligence Workshop — Examines unified multimodal architectures that go beyond next-token prediction for vision, language, audio, and action....
• Real-World Constrained Generative Models — Develops diffusion and flow-based generative models that satisfy real-world constraints and align with human preferences....
• AI with Recursive Self-Improvement — Explores the theoretical foundations and practical implications of AI systems that can improve their own...
• Scientific Methods for Understanding Deep Learning — Brings rigorous scientific methodology to understanding the mechanisms, generalization, and emergent behaviors of deep neural...
• Scaling Post-training for LLMs — Examines how RLHF, DPO, instruction tuning, and other post-training methods scale with compute, data, and...
• Scaling Post-Training for LLMs — Investigates RLHF, DPO, instruction tuning, and other post-training methods for scaling and aligning large language...
• Time Series in the Age of Large Models — Explores how large pre-trained models can be adapted for time series forecasting, classification, and anomaly...
• Time Series in the Age of Large Models — Brings together the time series forecasting and large model communities to examine how foundation models...
• Test-Time Updates — Investigates methods that allow models to adapt at inference time - including test-time training, in-context...
• Test-Time Updates Workshop — Studies how models can adapt and improve at inference time using test-time training, prompting, and...
• AI Verification in the Wild — Bridges formal verification and machine learning, exploring how to provide provable guarantees about model behavior...
• AI Verification in the Wild — Advances formal verification, testing, and certification of AI systems for safety-critical real-world deployments.
• World Models Workshop — Explores internal world models in AI systems for planning, imagination, and model-based reinforcement learning.