AI
AI agents can be powerful—and unpredictable. AgentGuardian is a new security framework that keeps them on-task by learning and enforcing access rules tailored to each agent. How it works: * Staging phase: It watches the agent run in a controlled setting, logging execution traces and typical inputs to learn what
AI
Most top video AIs are locked up. Molmo2 opens the door: open weights and open datasets, built to understand videos and ground that understanding by pointing to and tracking objects in the pixels. * Data you can build on: 7 new video datasets and 2 multi-image sets, including rich video captions,
AI
Smaller, smarter music AI Foundation models are powerful—but huge and pricey to run. This study shows how to keep music understanding strong while trimming model size and compute. * What’s new: A lean architecture that swaps heavy self‑attention for Branchformer + SummaryMixing, plus a simple random quantizer to learn
AI
New AI that keeps creativity: DPWriter LLMs trained with reinforcement learning often play it safe, shrinking the variety of their stories. DPWriter flips the script by planning before writing. It breaks generation into semi-structured steps, then uses Diverse Planning Branching to explore multiple, intentionally different routes. A group-aware diversity reward
AI
Large language models can ace math benchmarks yet still stumble on simple number sense because they treat numbers like ordinary words. This work fixes that by giving models a value-aware way to read numbers. How it works: whenever a number appears, the input is augmented with a tiny prefix token
AI
What if AI could think with pictures? Omni-R1 is a new multimodal AI that doesn’t just “talk through” problems—it draws its intermediate steps. Instead of relying on one fixed reasoning style, it unifies many skills (like zooming into regions, pointing to objects, or marking paths) by generating small
AI
Big business decisions often rely on complex optimization models, but building them is slow and manual. Meet LEAN-LLM-OPT, a lightweight, multi-agent AI that auto-formulates large-scale optimization models from a plain-English problem description and datasets. How it works: two planner agents design a step-by-step workflow for similar problems; a builder agent
Robotics
Humanoid robots need to coordinate physically with people—lifting, handing over, steadying—but we lack data of humans interacting with robots. What if they learned from humans interacting with humans? The catch: simply mapping human motions onto a robot often breaks crucial touches and supports. The team proposes PAIR (Physics-Aware
AI
AI discovers better ways to fast-charge batteries Charging batteries quickly without wearing them out is hard—and testing each new idea takes time and money. Researchers show that large language models (the tech behind chatbots) can help design smarter charging “recipes.” * Two approaches: Prompt-to-Optimizer (P2O), where an AI writes small
AI
TL;DR An LLM acts as a coach to prune another LLM, shrinking it ~45% while preserving key knowledge and accuracy. Traditional pruning uses fixed rules and often wipes out facts. This paper lets a foundation model adaptively choose which layers to trim each round. It reads layer sensitivity snapshots—
AI
AI is turbocharging factories, supply chains, and maintenance—but it also widens the ethical playing field. This chapter maps where industrial AI meets ethics: transparency, accountability, fairness, data sharing, and responsible R&D. Core message: building ethics into systems from day one accelerates innovation and trust. Ethics isn’t
AI
TL;DR Debiasing big language models is costly. This study shows compact “BabyLMs” can mimic how larger BERT-style models learn biases—so researchers can test ideas faster and cheaper. * BabyLMs (small, BERT-like models on tiny, editable corpora) track the same bias and performance patterns as standard BERTs. * Correlations hold across