Agentic University (AU) is a human‑participatory multi‑agent system that uses large language models to provide personalized tutoring, content generation, and adaptive learning paths. Learners act as organization members and collaborate with specialized agents to create and revise a Capability Program that links personal goals, learning activities, and layered assessment. Evaluation shows that AU improves program organization, assessment design, and revision quality, yielding higher assessment gains and stronger cognitive support compared to baseline approaches.
The paper "rTaste: Do Research Agents Have Taste? Evaluating Early Stage Research Judgment" investigates whether advanced language models can act as research agents that propose and rank minimum viable experiments (MVEs) for recent NLP papers. Six frontier models were given seven papers and tasked with forecasting experiment outcomes and ranking them; a separate executor ran all MVEs and auditors evaluated the results. Of 415 MVEs, 194 were deemed sound, valuable, and new, but only 73 were judged beyond the seed paper and non‑obvious; while agents’ first choices matched auditors’ top picks more often than random, they did not outperform simple heuristics, and team models added cost without clear quality gains.
"whyItMatters":"The study provides a dataset and evaluation harness for assessing AI‑driven research planning, highlighting both the potential and current limitations of using language models as research agents."
The paper introduces KITScenes Aerial, an extension of the KITScenes Multimodal dataset that adds synchronized aerial imagery and high‑resolution neural representations to existing onboard sensor data, vehicle poses, and Lanelet2 map annotations. It evaluates mapping across all 53 KITScenes map classes, showing that aerial context significantly improves fine‑grained and extended‑range mapping, with higher‑resolution imagery boosting performance further, especially for challenging elements like traffic lights. The study also assesses cross‑view localization under geographically disjoint conditions, achieving a mean translation error of 5.13 m with a fine‑grained feature‑matching baseline.
The article describes how Jacob Sansbury’s team built a massively multiplayer game called Spawn, where players can prompt an in‑game AI agent, Savi, to modify the world—creating castles, dragons, and more. The development process is fully autonomous, with Fable‑powered agents that read errors, bug reports, and metrics, then directly modify live code and ship changes to production every hour. The system treats any hour without a ship as a severe incident, enabling rapid rollbacks and continuous deployment at a scale that costs over $3 million in AI tokens monthly.
The paper introduces EpiCoach, an agentic system that treats a single athlete as the subject of accountable N‑of‑1 science. It uses a mechanistic digital twin built from wearable data to make probabilistic predictions about future training sessions, and an evidence‑gated hypothesis registry records each claim and its outcomes. Over eighteen months, the system’s predictions were evaluated both retrospectively and prospectively, showing competitive performance and demonstrating a framework for auditable, falsifiable personalization in digital health.
Wizard is an agentic game world model that allows real‑time, precise control over character motions while generalizing zero‑shot across new scenes, characters, and actions. It builds on a dataset of 750 scenes and 178 characters, using skeleton and geometry representations as structured proxies to condition a video generation model and an adaptive masking mechanism to improve alignment. The framework explicitly maintains and updates game states, enabling players to freely configure virtual worlds and exercise fine‑grained control while preserving visual quality and gameplay consistency.
RFW‑A introduces a hybrid approach that combines agentic synthesis with procedural generation to create reusable category‑level generators for articulated assets. By testing and refining these generators across sampled configurations, the system can produce over 400 K high‑quality, simulation‑ready assets across 500+ categories at a fraction of the cost per asset. The resulting dataset demonstrates strong structural validity, robust physics simulation, and improves downstream tasks such as part‑level articulation estimation and zero‑shot policy transfer to real objects.
The paper introduces a hybrid framework combining Model Predictive Control (MPC) and Reinforcement Learning (RL) for autonomous highway driving of heavy‑duty trucks. A finite‑horizon MPC, solved via dynamic programming, uses a pre‑trained RL model to estimate terminal costs, addressing the limitations of short‑horizon MPC and the lack of constraint guarantees in pure RL. In simulation with reactive traffic, the hybrid approach achieves zero collisions while incurring the lowest cost among compared methods.
Ben Hylak envisions a product architecture where every tool is a headless micro‑service that feeds into an agent, freeing developers from underlying complexities. The focus is on delivering a superior user interface that customers love, while the internal workings remain abstracted. This approach aims to combine powerful backend functionality with a delightful front‑end experience.
The article discusses how AI agents are transforming software development by autonomously handling tasks across the SDLC, yet their actions are fleeting and lack traceability. It highlights the emerging security risks—intent mismatch, injection, and supply‑chain vulnerabilities—and the absence of standards for agentic provenance and governance. To address this, the authors propose Agentic Process Evidence (APE), a signed, tamper‑evident, machine‑readable attestation that links AI agent sessions to outcomes, enabling policy‑as‑code control and human oversight only when necessary.
AI Cosmos is an agentic infrastructure designed to enhance situated learning by organizing learners and specialized agents within scenario-based learning environments. It integrates task-relevant knowledge with executable workbenches and uses an asset‑memory loop to preserve practice outcomes as reusable assets with provenance. In two randomized controlled studies, AI Cosmos outperformed conventional LLM approaches in legal reasoning and achieved higher mastery of unseen programming problems compared to an LLM‑as‑Teacher baseline.
WorldProbe is a new diagnostic benchmark that tests video world models by letting a goal‑conditioned multimodal agent adapt camera actions online during generation, creating 164 test cases across four probing families. The benchmark evaluates 10 state‑of‑the‑art models over more than 1,600 rollouts, totaling roughly 800 minutes and horizons up to 180 seconds, revealing that while local visual and content quality can be maintained for longer, long‑range memory and evolution remain weak. The study highlights a gap between locally stable generation and persistent world consistency that existing short‑interaction benchmarks miss.
The paper evaluates how large language models, when integrated into a harness that supplies information and executes software, perform as agents in acoustic engineering tasks. In one experiment the agent accurately detects and locates faults in vehicle interior audio recordings, showing that a weak model limits performance regardless of harness design. In a second case study the agent reliably designs metamaterial unit cells to meet specified insertion loss criteria, illustrating how different harness components influence agent behavior.
The paper presents EpidemIA, a training‑free AI Multi Agent System that lets clinicians and administrators use epidemiological models and perform statistical analyses through natural language conversations. The system runs on a single‑GPU workstation, employs a locally hosted open‑weights LLM to keep data private, and includes a modular suite of expert‑written models. Using Emergency Department visit data from IRCCS Sant’Orsola, the authors demonstrate that users can conduct both simple statistical queries and complex dynamical predictions via natural language.
EditAgent rethinks agentic image editing by expanding both the action and observation spaces. It routes each editing step to general editors and specialized methods, attaching verification records that allow off‑the‑shelf VLM judges to outperform finetuned reward models. The approach yields parallel candidate generation and a single listwise judgment, which experts prefer over sequential refinement in 78.5% of cases and improves two open‑weight editors across three benchmarks.
OpenAI employs layered monitoring to identify misalignment in frontier models and detect unauthorized agent behavior. This approach allows responders to intervene before incidents worsen. The system focuses on early detection and prevention of potential risks.
The paper introduces AgenticDS, a contract-governed agentic framework that transforms a training data query into a typed search plan and executes it through deterministic operators, ensuring that retrieved datasets meet specified constraints. By verifying provenance‑grounded evidence before ranking, AgenticDS outperforms existing hybrid retrieval and commercial LLM agents, improving NDCG@10 by 25.7% on the HuDS‑Bench benchmark. A controlled search‑to‑train study demonstrates that the datasets found by AgenticDS lead to superior downstream accuracy in mathematics, code, and medicine tasks.
The paper introduces SkillCompiler, a framework that transforms self‑evolving agent skills into executable programs to reduce uncertainty from repeated LLM reasoning. It decomposes skills into fine‑grained operations, selectively compiles deterministic parts while retaining LLM reasoning for context‑dependent steps, and reuses validated programs across evolving skills. Experiments on five benchmarks and six model families show SkillCompiler outperforms state‑of‑the‑art methods, improving task performance by up to 32.33%.
The paper investigates whether scaling toward artificial general intelligence (AGI) can be achieved by co‑designing a lightweight, semantically aware predictor (System‑1) with a resource‑intensive executive (System‑2). It introduces a conditional interaction loop where System‑1 estimates the utility of candidate actions, System‑2 executes those that pass an adaptive threshold, and the outcomes update System‑1. The study shows that, under ideal conditions, this approach can reduce execution compute by a factor related to the density of utility‑relevant interactions, and it proposes that this density is a key metric for scaling efficiency.
Youtu-Parsing-Omni is an omni‑modal parser that transforms documents, images, audio, and video into structured records grounded in space and time, enabling agents to read, locate, and cite evidence across modalities. It uses a single encoder—initialized from a pre‑trained language model—to encode all modalities with shared weights in a unified space‑time coordinate system, while a few fusion layers allow video frames to attend to concurrent audio for cross‑modal fusion and temporal alignment. The system unifies ten parsing tasks under one schema, employing a self‑evolving post‑training procedure called OmniSchema‑Aware On‑Policy Distillation (OSAD) that co‑evolves the parser and its supervision, achieving state‑of‑the‑art performance on OmniDocBench v1.6 and leading open‑weight models on OmniParsingBench, with strong results on audio, video, chart‑parsing, and chemical‑fingerprint agreement.