Papers (1699)

Filter: Sort:

Prompting LLM agents to be less or more 'agreeable' changes how they communicate but only matters for unstructured tasks: low agreeableness substantially worsens outcomes in research-style collaboration and bargaining, while coding milestone completion is largely unchanged.

medium quasi_experimental Full text arxiv

When Does Personality Composition Matter for Multi-Agent LLM Teams? · Aryan Keluskar, Amrita Bhattacharjee, Huan Liu · Jun 25, 2026

Combining many LLMs yields limited upside: ensemble accuracy is capped by the rate at which every model fails, and across 67 models observed co-failure rates are 2–3 times larger than common Gaussian-copula models predict. As a result, simple voting or mixture strategies seldom beat the best individual model unless a reliable per-query router is available.

medium correlational Full text arxiv

When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models · Josef Chen · Jun 25, 2026

Subtle self-promotional text can game LLM résumé screeners when few applicants do it and candidates look similar, but the effect evaporates once manipulation is common; occasional rank flips allow weaker applicants to leapfrog stronger ones, creating fairness risks.

medium rct Full text arxiv

Prompt Injection in Automated Résumé Screening with Large Language Models: Single and Multi-Injection Settings · Preet Baxi, Jiannan Xu, Jane Yi Jiang, Stefanus Jasin · Jun 25, 2026

Open-source generative models have democratized online 'nudification': a study of an anonymous community found 24,105 AI-generated non-consensual sexual images and videos, with non-celebrities now comprising 55.8% of targets and a handful of prolific users sourcing and sharing the models and tutorials that sustain the ecosystem.

medium descriptive Full text arxiv

From Celebrities to Anyone: Characterizing AI Nudification Content, Technology, and Community Dynamics on 4chan · Chi Cui, Yixin Wu, Yang Zhang · Jun 25, 2026

Making agents open-source can make selfish learning cooperative: in parametric open-source games the authors prove an exact coupling threshold where gradient-ascent learners flip from defection to cooperation, and show that a neural-semantics sensitivity ratio predicts when shared parameter access steers outcomes toward cooperation.

n/a theoretical Full text arxiv

Parametric Open Source Games · Aleksandar Todorov, Jesse ten Napel, Alexander Müller · Jun 25, 2026

Agentic AI use has surged: active Codex users quintupled in the first half of 2026 and multi-agent, skill-based workflows are increasingly common; internal OpenAI staff report large token-output gains, signalling substantial workplace change even as external adoption remains uneven.

medium descriptive Full text arxiv

The Shift to Agentic AI: Evidence from Codex · Drew Johnston, David Holtz, Alex Martin Richmond, Christopher Ong, Prasanna Tamb... · Jun 25, 2026

Agent configurations for LLM-driven tooling are widely copied and under-governed: 10.1% of tracked paths are exact duplicates across independent repositories and under 1% declare permissions. The paper proposes Rel(AI)Build, a deterministic, content-addressed control plane to audit, lock, and enforce permission and supply-chain invariants before agent execution.

medium descriptive Full text arxiv

A Deterministic Control Plane for LLM Coding Agents · Padmaraj Madatha · Jun 25, 2026

Combining many LLMs and multiple sampled outputs dramatically boosts effective performance: across 21 models and 16 benchmarks, multi-model multi-generation selection reduces error rates by up to 82% versus typical single-model single-run evaluation and reaches state-of-the-art accuracy at 85% lower cost.

medium descriptive Full text arxiv

The Capability Frontier: Benchmarks Miss 82% of Model Performance · Bradley Fowler, Ryan Smith, Daniel Thi Graviet, William Myers, Joshua Greaves, N... · Jun 25, 2026

An LLM-driven pricing agent lifted gross merchandise value by 13.2% and improved ROI by 7.6% in a two-week randomized trial on Tao Factory, delivering interpretable, long-term-value-aligned price recommendations.

medium rct Full text arxiv

AIGP: An LLM-Based Framework for Long-Term Value Alignment in E-Commerce Pricing · Chennan Ma, Yanning Zhang, Siqi Hong, Xiuchong Wang, Fei Xiao, Keping Yang · Jun 25, 2026

AI is not an automatic growth engine: its economic payoff depends on human capital, data infrastructure and institutional complements, causing similar AI investments to yield divergent outcomes across countries and firms.

medium review_meta Summary openalex

The Impact of Artificial Intelligence as a General-Purpose Technology on Economic Growth and Structural Transformation: An Innovation Ecosystem Perspective · Sultan Salur Kucuk · Jun 25, 2026

AI industry clusters first undermine but eventually enhance China's ability to jointly cut agricultural pollution and carbon emissions, producing a clear U-shaped effect; technological progress mediates the turning point and impacts are stronger where marketization and industrialization are higher.

low correlational Summary openalex

How Does Artificial Intelligence Industry Agglomeration Affect Agricultural Pollution–Carbon Reduction Synergy in China? Evidence from a Marginal Cost Perspective · Shuang Gao, Dan Li, Masaaki Yamada, Haisong Nie · Jun 25, 2026

A governance model that lets AI agents plan but not execute high‑risk actions: execution requires independent authoritative attestations cryptographically bound to declared intents and deterministic policy checks, with tamper‑evident logging for re‑verification. The approach promises institutional control without micromanaging agent reasoning, but it lacks real‑world validation and depends on attester infrastructure and legal acceptance.

n/a theoretical Full text arxiv

Governing Actions, Not Agents: Institutional Attestation as a Governance Model for Autonomous AI Systems · Jakob Salfeld-Nebgen · Jun 24, 2026

The ERC-8004 trust layer for autonomous AI agents records many identities but few verifiable services, and its reputation registry is easily gamed; after filtering coordinated Sybil activity, most rated agents have no trustworthy feedback.

medium descriptive Full text arxiv

Can Trustless Agents Be Trusted? An Empirical Study of the ERC-8004 Decentralized AI Agent Ecosystem · Xihan Xiong, Zelin Li, Wei Wei, Qin Wang, William Knottenbelt, Zhipeng Wang · Jun 24, 2026

How AI is placed in lessons decides whether it teaches or short-circuits learning — an unguarded AI helper reduced students' unaided exam scores by about 17%, while a constrained assistant removed the harm and a tutor-style system roughly doubled learning. Educators should require a hard initial attempt and an unaided final check, using guarded AI scaffolds in between.

medium commentary Full text arxiv

The Effortless Trap: Productive Struggle, AI, and the Illusion of Learning · Mario Brcic, Stjepan Frljic · Jun 24, 2026

AI is automating routine developer work and reshaping roles inside SAP's Business Technology Platform; expert interviews and a 24-person workshop reveal growing human–AI collaboration and urgent need to update role taxonomies and governance for AI-native enterprise systems.

low descriptive Full text arxiv

The impact of artificial intelligence on enterprise software user roles · Isabel Unger, Elizangela Valarini, Martin Schrepp, Nina Hollender, Gabriela Roch... · Jun 24, 2026

AI boosts efficiency at financial firms but creates new systemic risks: common data and model architectures produce algorithmic homogeneity and opaque decision chains that amplify shocks into digital herds and 'algorithmic resonance'. Macroprudential frameworks must be redesigned for an era where technological dependence, not just leverage, drives fragility.

n/a theoretical Full text openalex

A Theoretical Framework for AI and Financial Stability: The Tension Between Micro-Efficiency and Macro-Risk · Tianle Chang · Jun 24, 2026

Firms that bring in new talent build more AI capacity: Chinese A‑share companies in Shanghai and Shenzhen that introduce external talent show significantly higher AI development, an effect robust to IV estimation. The uplift appears to operate via improved financing access and workforce quality and is strongest in manufacturing and specific regions.

medium quasi_experimental Full text openalex

The Impact of Talent Introduction Intensity on Corporate Artificial Intelligence Levels: Empirical Evidence from Chinese A-Share Listed Companies · Shanlin Bi · Jun 24, 2026

AI adoption accelerates value‑chain upgrading in China’s equipment manufacturing by boosting human capital and product R&D; gains are strongest in eastern and western provinces and in capital‑ and technology‑intensive subsectors, while labor‑intensive segments experience negative effects.

medium quasi_experimental Full text openalex

The impact of artificial intelligence on value chain upgrading in China’s equipment manufacturing industry · Hui Li, Yinzhong Chen, Jingfeng Huang · Jun 24, 2026

Individual data-rights regimes leave a collective accountability gap in algorithmic workplace governance; a new 'Collective Algorithmic Rights' framework of five institutional levers could restore collective bargaining power and oversight, though capacity to adopt it varies sharply across the EU, US, Latin America, Asia and Türkiye.

n/a theoretical Full text openalex

COLLECTIVE ALGORITHMIC RIGHTS: A NEW RIGHTS ARCHITECTURE FOR GLOBAL DIGITAL WORK REGIMES · Emrulah Tekin · Jun 24, 2026

Occupations most exposed to AI pay a premium rather than a penalty, with the wage advantage concentrated among cognitively demanding roles; the pattern holds across the wage distribution and in instrumental-variable tests.

medium quasi_experimental Summary openalex

Artificial intelligence exposure and occupational wages: Evidence from the United States · Ozan Atalay · Jun 24, 2026

Real-estate recommendation research has fallen behind modern AI: only 15% of studies use deep learning and none adopt state-of-the-art explainability or transfer-learning tools. Eighty percent rely on proprietary data, most work targets Asian residential markets, and fairness audits are rare—prompting authors to propose a trust-focused evaluation framework to steer future systems.

n/a review_meta Summary openalex

Real Estate Recommender Systems: A PRISMA-Compliant Systematic Review of Multimodal, Spatio-Temporal, Explainable, and Fairness-Aware Innovations · Musa Mbedzi, Thulane Paepae · Jun 24, 2026

Expanding short-selling access in China’s stock market nudges firms toward AI: companies exposed to the short-selling mechanism raise AI-related investment and patenting, partly driven by increased R&D spending.

medium quasi_experimental Full text openalex

Research on the Impact of Short Selling System on Enterprise Artificial Intelligence Application · Ruihang Liu · Jun 24, 2026

Human–AI collaboration roughly halves the time required to reproduce scientific code in a randomized trial, even after benchmark accuracy has saturated; improving the benchmark and adding an OOD suite reveals construct-validity and robustness issues that accuracy alone would miss.

medium rct Full text arxiv

Life After Benchmark Saturation: A Case Study of CORE-Bench · Nitya Nadgir, Sayash Kapoor, Kangheng Liu, Peter Kirgis, Matilda Orona, Stephan ... · Jun 23, 2026

Augmented-reality feedback can sharply boost human-robot picking efficiency, but only if overlays avoid blocking the real world; an obstruction-aware AR system delivered a 46% speedup overall and 61% on visibility-dependent subtasks in a lab warehouse simulation.

medium quasi_experimental Full text arxiv

ARTOO-DARTU: Studying AR-HRC With AR Obstruction Mitigation During a Warehouse Task · Christian Fronk, Hanting Ye, Zhehan Qu, Maria Gorlatova · Jun 23, 2026

LLM assistants often convince developers they've met HIPAA non-functional requirements, yet expert-verified accuracy is low; verbose, information-dense replies hurt user satisfaction while proactive behavior helps it.

medium correlational Full text arxiv

Accuracy and Satisfaction in Multi-Turn LLM Dialogues for NFR Assessment · Ali Pourghasemi Fatideh, Wilder Baldwin, Maria Dhakal, Collin McMillan, Sepideh ... · Jun 23, 2026

« Prev 1 2 3 4 5 6 … 67 68 Next »