ICLR 2026 — всё! А мы не договорили — 5 мая 2026 г. в 15:38:00.374
ICLR 2026 — всё! А мы не договорили Международная конференция по обучению представлений завершилась 27 апреля, и пока наши исследователи возвращаются из Бразилии, дорассказываем об участии AIRI на воркшопах — представили на них 20 работ: Machine Learning for Genomics Explorations (MLGenX): • ️Modern Gene Finders: ab initio gene discovery benchmark with DNA language models ReALM-GEN: Real-World Constrained and Preference-Aligned Flow- and Diffusion-based Generative Models: • ️Guided Star-Shaped Masked Diffusion Deep Generative Model in Machine Learning: Theory, Principle and Efficacy (2nd Workshop): • ️Discrete Bridges for Mutual Information Estimation • ️Inverse Entropic Optimal Transport Solves Semi-supervised Learning via Data Likelihood Maximization • ️Overclocking Electrostatic Generative Models • ️Inverse-distilled Diffusion Language Models • ️Time-Correlated Video Bridge Matching • ️FMMI: Flow Matching Mutual Information Estimation • ️One-Step Residual Shifting Diffusion for Image Super-Resolution via Distillation Agentic AI in the Wild: From Hallucinations to Reliable Autonomy: • ️Efficient Hallucination Detection for LLMs Using Uncertainty-Aware Attention Heads • ️Steering Large Language Models Toward Clarification through Sparse Autoencoders Representational Alignment Workshop: • ️OrtSAE: Orthogonal Sparse Autoencoders Uncover Atomic Features Generative AI in Genomics (Gen^2): Barriers and Frontiers: • ️Back to BERT in 2026: ModernGENA as a Strong, Efficient Baseline for DNA Foundation Models • ️Modern Gene Finders: ab initio gene discovery benchmark with DNA language models Principled Design for Trustworthy AI: Interpretability, Robustness, and Safety Across Modalities: • ️The Rogue Scalpel: Activation Steering Compromises LLM Safety VerifAI-2: The Second Workshop on AI Verification in the Wild: • The Dual Nature of Unlearning: Impact of Fact Salience and Model Fine-Tuning NFAM Workshop: New Frontiers in Associative Memories: • ️Dynamics of modern Hopfield networks • ️GradMem: Learning to Write Context into Memory with Test-Time Gradient Descent • ️Extending LLM Context via Associative Recurrent Memory Learning Meaningful Representations of Life (LMRL): • ️Physics-Constrained Correlation-Aware Attention for Collective Cell Dynamics И делимся второй порцией фотографий из Рио! #AIRIнаКонфе