Autonomous Agents

Opening — Why this matters now Everyone wants AI agents that can plan, reason, and execute multi-step work. Fewer people ask the impolite question: Can they keep doing it when the task gets longer? A new ICLR 2026 paper studies this with unusual discipline. Instead of another benchmark made of messy internet text and leaderboard optimism, the authors use shortest-path planning in synthetic maps to isolate one brutal truth: many models can transfer skills to new environments, yet still collapse when the sequence of decisions extends too far. ...

Opening — Why this matters now Everyone wants AI to grade AI. It is faster, cheaper, and does not ask for lunch breaks. From summarization benchmarks to model leaderboards, LLM-as-judge systems now sit quietly inside many evaluation pipelines, handing out scores with bureaucratic confidence. There is only one minor complication: no one has been checking whether the judge is reliable on any given case. ...

Opening — Why this matters now Everyone wants AI that can evaluate AI. It is cheaper than humans, faster than humans, and—according to many slide decks—more scalable than reality itself. Modern AI pipelines increasingly rely on LLM-as-a-judge systems to rate safety, quality, policy compliance, and readiness for deployment. These judges decide whether a model is helpful, harmful, safe, or suspect. Conveniently, they do so without lunch breaks. ...

Opening — Why this matters now Everyone wants AI in the real world: warehouse robots, smart glasses, autonomous carts, industrial copilots, eldercare devices. Unfortunately, the real world insists on being noisy, dark, shaky, delayed, expensive, and occasionally ridiculous. Most modern AI systems were designed for clean, pre-captured data and abundant compute. Physical AI gets none of those luxuries. A blurry camera frame cannot be reasoned into sharpness by sheer optimism. A dead battery does not care how many parameters your model has. ...

Opening — Why this matters now Electric grids are becoming less predictable, more distributed, and less forgiving. Renewables fluctuate, demand spikes move faster, and operators must make decisions across sprawling networks under hard physical constraints. Meanwhile, everyone would like AI to optimize infrastructure—preferably yesterday. There is one awkward detail: power grids are not ad-click systems. When recommendation engines fail, users get odd suggestions. When grid control fails, cities get darkness. ...

Opening — Why this matters now Everyone wants autonomous coding agents. Fewer people ask the less glamorous question: how do they remember? Most current agents solve tasks as if each assignment is a surprise party. They may retain notes from similar prior tasks, but usually only within the same benchmark or domain. That is tidy for research papers and terribly unrealistic for business operations. ...

Opening — Why this matters now Every industry has a bottleneck disguised as tradition. In academia, it is peer review: noble in theory, overloaded in practice, and increasingly powered by caffeine and resentment. The paper AI-Assisted Peer Review at Scale: The AAAI-26 AI Review Pilot reports something more consequential than a conference experiment. It documents a live deployment where 22,977 submissions each received an official AI-generated review in under 24 hours. No sandbox. No toy benchmark. Real papers, real authors, real consequences. ...

Opening — Why this matters now Everyone wants AI that can reason. Preferably about things that matter: machinery, logistics, engineering diagrams, medical imaging, factory operations. Unfortunately, many systems marketed as “reasoning models” are still glorified pattern matchers with a flair for confident prose. This paper, Reward Design for Physical Reasoning in Vision-Language Models, asks a sharper question: if we reward an AI differently, what kind of reasoning behavior do we get? The answer is refreshingly inconvenient. There is no universal reward signal that makes models smarter. There are only trade-offs, incentives, and consequences. Rather like management. ...

Opening — Why this matters now Everyone wants custom AI. Few want the invoices, GPU queues, brittle data pipelines, and endless hyperparameter arguments required to build it. Fine-tuning large language models remains one of the least glamorous bottlenecks in modern AI deployment. It is expensive, iterative, and strangely dependent on whoever in the room has the strongest opinions. ...

Opening — Why this matters now AI agents are graduating from chat windows into operational systems. They now book meetings, write code, reconcile spreadsheets, and increasingly, manipulate the physical logic of maps. That last category matters more than it sounds. Spatial decisions shape flood planning, logistics routes, emergency response, land use, insurance risk, and infrastructure spend. ...