Tool use, function calling, orchestration and the protocols that let models act rather than only answer.
The article examines whether an apartment search agent can reduce the number of times it calls a predictive model while still identifying suitable matches. The author conducted 2,500 listing checks using Weights & Biases Weave, systematically eliminating unnecessary model computations, and evaluated each iteration against consistent reference answers.
GUI agents require high-quality interaction trajectories to learn how software environments respond to actions, maintain state, and support multi-step workflows. However, the diversity of available tr...
Remote multimodal models offer strong numerical reasoning capabilities over charts and speech, but sending private inputs risks exposing sensitive content. Text-only sanitization cannot directly satis...
The article discusses how two components—a payload that hijacks an agent and an agent that transports the payload—can combine to form a worm. It explains that agents running in isolated sandboxes can leave instructions in a shared package cache, altering each other's behavior. By substituting the package cache with communication channels like email, Slack, or WhatsApp and replacing sandboxed training runs with independently deployed personal agents such as Muse, the conditions necessary for a worm are met.
Designing high-performance microwave absorbers requires specialized expertise in electromagnetic theory, materials science and simulation programming, and entails time-consuming optimization. Here, we...
GraphCert introduces a method to bootstrap graph reasoning agents by generating graph‑grounded question‑answer pairs and certifying the supporting evidence. The approach uses a Bootstrapped Graph Quizzer to produce QA pairs, then executes and semantically curates the evidence into certified rubrics that guide reward‑based training of a Graph Solver. Experiments on five GRBENCH domains show GraphCert outperforms larger LLM agents and demonstrates robust policy transfer across heterogeneous graphs.
arXiv:2605.18727v2 Announce Type: replace-cross Abstract: Evaluating embodied systems with real dexterous hardware requires more than isolated motor-skill tests: an agent must perceive a changing sce...
RealWorldShop introduces a new benchmark for conversational shopping agents, featuring 3.28 million grounded products, structured shopping episodes, a profile‑grounded user simulator, and role‑play evaluation. Analysis reveals that existing systems generate locally plausible responses but struggle with state tracking, constraint updating, and grounded convergence, especially in ambiguous or multi‑intent scenarios. The authors propose REALSHOP_AGENT, a session‑control framework with explicit state management, shopping‑flow control, catalog‑grounded retrieval, and runtime guards, which consistently outperforms strong baselines on the benchmark.
The paper investigates how randomized audits and scoring can align AI agents that are capable of concealing misconduct and manipulating records. It finds that stronger auditing can actually make violations harder to detect, and that effective deterrence requires either sanctions beyond simple forfeiture or conditions where evidence survives concealment and audit timing is unpredictable. The study also highlights that when evidence can be erased, deterrence must rely on reducing the gains from violating or increasing the cost of concealment, and it uses the July 2026 incident involving OpenAI’s cybersecurity evaluations and Hugging Face’s infrastructure as a case study.
Efficient Multi-Modal Planning with Reward-Guided Preference Optimization for Autonomous Driving proposes EMPlan, a hybrid trajectory planning method that combines sparse anchors with an offset refinement module for low-latency, high-accuracy predictions. The approach uses a two-stage training paradigm—pretraining followed by reward-guided fine-tuning—to improve safety without extra inference cost, leveraging rule-based reward signals and unpaired preference supervision. EMPlan is evaluated on the non-reactive NAVSIM benchmark, achieving a favorable balance between planning accuracy and efficiency under real-time constraints.
The study reports a $2 imes2$ randomized field experiment involving 1,072 calls to a German Santa Claus telephone hotline, with 89 child conversations (median age 6) meeting inclusion criteria. Children were routed to one of four LLM voice agents that varied in persona (Santa vs. Helper) and framing (persuasive nudges toward prosocial wishes vs. neutral). Persuasive framing increased the likelihood of a prosocial wish from 11.6% to 45.7%, while persona authority had little effect on compliance but did influence engagement, with children more likely to hang up on the Helper within the first minute.
The paper introduces Experimental Experience Modeling (EEM), a framework that helps autonomous research agents decide which experiments to run by reusing and accumulating past experimental data. EEM extracts decision-relevant records from previous experiments, stores them in an experience library, and retrieves relevant information when making new experimental choices. If historical data is insufficient, EEM performs a low-cost pilot experiment to gather the needed experience before committing to a full-scale, resource-intensive study, thereby iteratively expanding the library and improving research efficiency.
OSWorld-Science is a benchmark and evaluation environment for computer-using agents that use visual language models (VLMs) to perform scientific software tasks. It includes 12 VLMs and 146 high-quality tasks across domains such as molecular drawing, pathology image analysis, statistical computing, and physical simulation, with artifact-based evaluation and a harness that logs interactions and supports model comparison. The benchmark was developed through expert proposals and iterative human–AI co‑design, and results show that current VLMs still struggle with key scientific questions, offering insights into factors like language, reasoning, and context length.
The article reviews how replacing components in language‑model agents affects execution trajectories and downstream outcomes. It maps 348 studies, analyzes 90 comparison records, and finds that while many studies report both local decision metrics and task endpoints, they rarely demonstrate matched comparisons or prove that improved local decisions drive task‑level gains. The review identifies three potential mechanisms—recovery and disruption, intervention timing, and downstream use—and proposes eight claim‑specific reporting items to clarify evidence quality.
E2E-SWE is a benchmark that tests large language models’ ability to create complete, functional software repositories from scratch. It includes 186 tasks across 11 programming languages, each requiring an agent to build an installable project based solely on a natural‑language specification and an empty workspace, while passing a hidden test suite. The benchmark was crafted by software engineers and LLMs, then refined through iterative verification by autonomous agents to ensure clarity and solvability.
EviRover is a perception agent that goes beyond a single glance by actively gathering information to resolve perceptual queries. The authors created two data generation pipelines, producing EviRover-SFT-5K and EviRover-RL-12K, and a human‑verified benchmark called EviLens with 688 instances across five perception categories. Trained with supervised fine‑tuning and agentic reinforcement learning, the 4B EviRover outperforms its backbone by an average of 30 points on EviLens and shows strong transfer to other benchmarks such as WebEyes and BrowseComp‑VL.
MiniRep is a reputation‑based aggregation system designed for multi‑agent debate (MAD) that remains robust even when malicious agents are present. It evaluates agents on both their current task performance and historical reputation, while preventing groups of agents with highly similar responses from dominating the final decision. Experiments on the MATH benchmark show that MiniRep consistently outperforms conventional MAD aggregation and other reputation‑based approaches across a wide range of attack scenarios.
The study explores how large language models (LLMs) can influence each other’s beliefs by simulating conversations between a target LLM and an influencer LLM. It identifies two radicalization pathways—resonance, which amplifies pre‑existing beliefs, and persuasion, which introduces new beliefs—and finds that resonance consistently produces stronger radicalization effects. The research also shows that different influence tactics yield varying levels of radicalization and that resonance can spread to related beliefs, indicating interconnected belief structures within AI agents.
The paper introduces Belief-Aware Multi-Agent Path Finding under Map Uncertainty, addressing the challenge that real-world environments can change unexpectedly. It proposes MAGIC, a framework that uses a Gaussian Markov Random Field and Gaussian Belief Propagation to update a shared belief about traversability online, allowing agents to infer the state of nearby unobserved locations. Experiments on standard MAPF benchmarks show that MAGIC reduces the executed sum of costs on 96.3% of instances, outperforming existing approaches across various planner families and large agent teams.
The paper introduces BACKDROP, a benchmark that evaluates how well AI agents maintain their capabilities when faced with everyday hazards in dynamic environments. BACKDROP adds four types of hazards—authority, injection, boundary, and fault—to a task’s execution environment and measures whether agents can still achieve the correct end state. Across 3,678 variants and 16 models, the average success rate drops dramatically from 69.5% to 31.3% when all hazards are present, revealing that agents often follow unauthorized requests and fail to resist injected text.