1167 papers
This collection explores the diverse scientific landscape spanning from the elements of Cesium to the complex structures of Macromolecules. It captures a wide spectrum of inquiry, ranging from the behavior of specific atoms in physics and chemistry to the intricate study of large biological molecules that drive life processes. These papers reveal how fundamental components interact to shape materials, medicines, and the very building blocks of our world.
Gist.Science continuously monitors arXiv to bring you every new preprint in this rapidly evolving category. Our team processes each submission as soon as it is published, offering both plain-language explanations for general readers and detailed technical summaries for specialists. This dual approach ensures that groundbreaking discoveries are not just available, but truly understandable.
Below are the latest papers covering these essential topics, curated directly from the most recent uploads on arXiv.
This paper proposes an automated, multi-agent red-teaming framework that synthesizes challenging adversarial examples through iterative hypothesis generation and mutation to significantly enhance multimodal large language model robustness against safety violations without requiring human labeling.
This paper proposes and evaluates a stochastic filtering protocol () inspired by -priority sampling to mitigate double-counting bias in anonymous quorum sensing for robot swarms, demonstrating that while it improves estimate stability compared to baseline and randomized variants, it incurs a trade-off of slower recovery from errors.
This paper establishes that context engineering quality is a primary, measurable predictor of AI agent reliability, demonstrating through the ProofAgent-Harness that specific context criteria—such as grounding sufficiency and guardrail coverage—directly determine behavioral outcomes like hallucination resistance and instruction following, independent of the underlying language model.
This paper establishes that truthful delegation to automated proxies is optimal if and only if the proxy minimizes "within-range regret," revealing a fundamental trilemma where no safety guardrail can simultaneously be binding, truthful, and capability-preserving, thereby explaining the incentive for prompt engineering and jailbreaking.
This paper establishes a formal framework for analyzing disrupted urban mobility systems through a dynamic multi-layer equilibrium attractor, deriving conditions under which lower-dimensional indicator projections faithfully preserve the system's structural properties and decision-making power while providing a principled basis for city authorities to determine essential monitoring indicators.
This paper extends the Quality of Control (QoC) framework to practical robotic models by modeling end-to-end network effects on closed-loop performance, systematically analyzing the impact of control parameters on network latency and reliability, and experimentally validating these findings on a private 5G testbed to demonstrate that RELIABLE QoS policies significantly outperform BEST-EFFORT alternatives in dynamic collaborative navigation scenarios.
This paper argues that while randomized routing strategies for competing CAV fleets can outperform traditional optimal routing in diverse traffic environments, the market design must be modified to balance market-share incentives with system-wide travel time goals to prevent antisocial behavior and ensure social welfare.
This paper presents a multi-agent framework that combines fine-tuning, preference optimization, and retrieval-augmented generation to simulate partisan political coalition negotiations with ideological fidelity, introducing a novel lineage tracing system to validate how manifesto-anchored content predicts real-world agreement outcomes.
This paper proposes a theoretical framework to analyze the expressivity of multi-agent reasoning systems, deriving bounds on agent count, communication requirements, and speedups for specific algorithmic tasks, and validates these theoretical tradeoffs with empirical experiments on large language models.
The paper introduces MASPRM, a multi-agent system process reward model trained on terminal outcomes to score intermediate agent messages, which significantly improves inference-time search performance on reasoning benchmarks compared to existing models by enabling more effective step-level beam search and Monte Carlo Tree Search.