The Commonplace
Home Papers Evidence Explore Trends Syntheses Digests About 🎲 Workforce Futures

Papers (1699)

Filter: Sort:
Employees who perceive their employer is adopting AI report greater turnover intentions driven by identity threat; secure attachment dampens this effect while anxious attachment intensifies it.
low correlational Full text openalex
The impact of organizational artificial intelligence adoption on employee turnover intention: an identity-based perspective · Aiqing Wang, Weiping He · Jun 12, 2026
AI innovation and inclusive finance are linked to improved environmental carrying capacity in G7 economies, while globalization and urbanisation exert downward pressure; the results support a nonlinear income–environment (Load Capacity) relationship.
medium correlational Summary openalex
Artificial Intelligence, Financial Access, and the Path to Sustainable Development: Insights from G-7 Countries · Mohammad Ridwan, Kabaly P. Subramanian, Jaheer Mukthar K.P., Binata Rani Sen, Mi... · Jun 12, 2026
Automation is eroding income-tax receipts and increasing fiscal pressure, and a scoping review finds that a 'robot tax' could recoup revenue and finance retraining to soften worker displacement; the proposal, however, is conceptual and rests on heterogeneous, untested evidence.
low review_meta Full text openalex
Robot taxation as a fiscal policy instrument for sustainable employment: A scoping review with implications for developing economies · Chanelle Kistnasamy, Kiran Baldavoo · Jun 12, 2026
Intelligent-manufacturing pilots in China raise firm-level green innovation by boosting adaptive, absorptive and innovation capabilities, and stronger intellectual-property protection magnifies these gains.
medium quasi_experimental Summary openalex
Intelligent Manufacturing Dynamic Capabilities and Corporate Green Innovation: Empirical Evidence from China · Can Ding, Jianxin Xu, Jing Li · Jun 12, 2026
AI matters for innovation only when it becomes routine: firms with a digital culture that routinizes AI use convert AI investments into process, product and business-model innovation, while diffuse competence or isolated experimentation yields limited strategic payoff.
medium correlational Full text openalex
Vloga umetne inteligence in digitalne kulture pri spodbujanju digitalnih inovacij · Dejan Uršič · Jun 12, 2026
Generative AI use in agricultural firms is linked to more green innovation, largely because it strengthens firms' green dynamic capabilities; cohesive top management teams amplify both the direct and indirect effects.
low correlational Summary openalex
How Generative AI Applications Drive Green Innovation in Agricultural Enterprises: The Mediating Role of Green Dynamic Capabilities and the Moderating Role of TMT Behavioral Integration · X Y Li, Lei Xi · Jun 12, 2026
Shaping the execution environment, not just agent code, substantially increases LLM agents' ability to autonomously discover scientific solutions: EurekAgent's permission, artifact, budget, and human-in-loop engineering yields state-of-the-art results on multiple benchmarks and even finds a 26-circle packing solution for under $11 in API cost. The findings suggest that designing environments is a key lever for productive, low-cost autonomous research agents, though gains are shown on a narrow suite of tasks and depend on specific model/APIs.
medium descriptive Full text arxiv
EurekAgent: Agent Environment Engineering is All You Need For Autonomous Scientific Discovery · Amy Xin, Jiening Siow, Junjie Wang, Zijun Yao, Fanjin Zhang, Jian Song, Lei Hou,... · Jun 11, 2026
Synthetic data can support valid scientific inference when researchers can identify historical tasks that are exchangeable with the current task; the paper formalizes 'task exchangeability', gives provable correction methods, and demonstrates them on survey silicon samples and LLM autoraters.
n/a theoretical Full text arxiv
Valid Inference with Synthetic Data via Task Exchangeability · Lezhi Tan, Tijana Zrnic · Jun 11, 2026
A reinforcement-learning policy at DoorDash nudges the dispatch optimizer to increase batching and cut courier time costs without worsening delivery quality; the approach adapts objective weights via a constrained store-level multiplier learned from logged marketplace feedback.
medium quasi_experimental Full text arxiv
Multi-Agent Reinforcement Learning from Delayed Marketplace Feedback for Objective-Weight Adaptation in Three-Sided Dispatch · Haochen Wu, Yi Hou, Shiguang Xie · Jun 11, 2026
Training AI to mirror aggregate human values risks entrenching harmful political and social orders; instead, AI should be bound by a minimal floor of competence, factuality, honesty and lawfulness, with pluralism permitted only at surface conventions.
n/a commentary Full text arxiv
Position: Align AI to Our Aspirations, Not Our Flaws · Nikita Kazeev, Bui Nhat Huyen Phan · Jun 11, 2026
Next‑generation time‑series foundation models beat classical forecasters in zero‑shot prediction for cloud workloads, but they do not automatically improve server consolidation outcomes; choosing the right predictive quantile matters more for balancing efficiency and reliability.
medium descriptive Full text arxiv
CloudCons: A Comprehensive End-to-End Benchmark for Cloud Resource Consolidation · Xiaobin Zhang, Lefei Shen, Mouxiang Chen, Zhuo Li, Hongkai Li, Han Fu, Jianling ... · Jun 11, 2026
Nearly half of AI-generated code fixes are rejected, creating substantial review and validation waste; a qualitative analysis of 306 non-merged PRs identifies 14 failure reasons and recommends better guidance, constraints, and CI-aware validation to curb wasted effort.
medium descriptive Full text arxiv
Understanding the Rejection of Fixes Generated by Agentic Pull Requests -- Insights from the AIDev Dataset · Mahmoud Abujadallah, Ali Arabat, Mohammed Sayagh · Jun 11, 2026
Instruction files for code-writing AI are hit-or-miss: in an analysis of 15,549 agentic pull requests across 148 projects, only 27.7% of projects raised merge rates by at least 20% after adding instructions while 26.4% fell—projects that gained tended to use longer, more structured instructions.
medium quasi_experimental Full text arxiv
Toward Instructions-as-Code: Understanding the Impact of Instruction Files on Agentic Pull Requests · Ali Arabat, Mohammed Sayagh · Jun 11, 2026
Precomputing and hosting LLM KV caches can eliminate repeated prefill work and cut per-document prefill costs by tens of times: on Qwen3-4B a popular 3,774-token document served to millions could cost millions less to deliver, creating a lucrative provider-hosted CDN opportunity.
medium descriptive Full text arxiv
Can I Buy Your KV Cache? · Luoyuan Zhang · Jun 11, 2026
Generative AI reaches more privileged graduates in India: SC and ST graduates are 0.24–0.37 SD less likely than upper-caste peers to work in AI-exposed occupations within the same district. With AI exposure associated with wage premiums of up to 20%, the technology risks widening India's caste earnings gap.
medium correlational Full text arxiv
The Privilege of Exposure: Caste and Generative AI in India's Graduate Labour Market · Kaibalyapati Mishra · Jun 11, 2026
AI coding assistants didn’t raise architectural defects but they did bloat code: in 151 open-source Java projects, adopters saw no change in total architectural smells while code size grew ~12.8%, causing a 6.7% drop in smell density driven by larger codebases rather than cleaner architecture.
medium quasi_experimental Full text arxiv
Mining Architectural Quality Under Agentic AI Adoption: A Causal Study of Java Repositories · Oliver Aleksander Larsen, Mahyar T. Moghaddam · Jun 11, 2026
Brick, a multimodal router that scores models on six capability dimensions and estimates per-query difficulty, slightly outperforms the best single model (76.98% vs 75.02% accuracy) while letting operators trade quality for cost. At a neutral setting it preserves 74.11% accuracy at 4.71x lower cost and can cut costs up to 22.15x (with accuracy loss) while halving median latency.
n/a descriptive Full text arxiv
Brick: Spatial Capability Routing for the Mixture-of-Models (MoM) Paradigm · Francesco Massa, Marco Cristofanilli · Jun 11, 2026
Treating commercial CAD as COM-call programs, not pixel-control chores, dramatically improves autonomous task success: GUI-driven agents fail on industrial CAD, while COM execution plus a self-correcting agent delivers substantial, robust gains on a new real-world benchmark.
n/a descriptive Full text arxiv
ComAct: Reframing Professional Software Manipulation via COM-as-Action Paradigm · Jiaxin Ai, Tao Hu, Xuemeng Yang, Shu Zou, Hairong Zhang, Daocheng Fu, Yu Yang, H... · Jun 11, 2026
Large language models can autonomously breach simulated servers: tested across 300 targets, 19 models show penetration success rates from roughly 11% to nearly 70%, and capability gains track improved attack performance; however, results are based on simplified, controlled environments and may overstate real-world risk.
medium descriptive Full text arxiv
The Emergence of Autonomous Penetration Capabilities in Large Language Model-Powered AI Systems · Jiaqi Luo, Jiarun Dai, Zhile Chen, Jia Xu, Weibing Wang, Yawen Duan, Brian Tse, ... · Jun 11, 2026
Generative AI is remaking business education — accelerating ideation, case analysis and managerial skill development while exposing gaps in critical thinking, authorship, access and employability readiness. Universities and firms must redesign curricula, embed AI literacy and ethics, and adopt responsible policies to prepare graduates for AI-mediated organizations.
n/a review_meta Summary semantic_scholar
Instructing Higher Education in the Era of Generative AI: Implications for Managerial Decision-Making, Business Ethics, and Workforce Readiness · Ramya Devarajan · Fetched Jun 11, 2026
Machine learning mostly increases output by improving predictions; deep learning pushes automation and capital intensity in skilled sectors; and Generative AI is shifting value toward cognitive augmentation, reshaping employment in knowledge industries.
medium review_meta Summary semantic_scholar
AI Technologies and Economic Transformation: A Systematic Review Comparing Machine Learning, Deep Learning, and Generative Models · Lavanya Singla, Sushil Laddhu, Surbhi Tiwari, Shweta Goel · Fetched Jun 11, 2026
Automatically generated multi-agent systems cost up to ten times more but often perform worse than a single-agent Chain‑of‑Thought baseline; expert‑crafted multi‑agent architectures can beat generated ones on a synthetic decomposition task, exposing automated design's inefficient complexity.
medium descriptive Full text arxiv
The Illusion of Multi-Agent Advantage · Prathyusha Jwalapuram, Hehai Lin, Chuyuan Li, Fangkai Jiao, Sudong Wang, Yifei M... · Jun 11, 2026
A research harness for LLMs slashes AI-assisted research failures—from 72% down to 16%—by preventing models from executing data work and inserting human decision gates; deterministic computation and human checkpoints each improve reliability and together appear complementary.
medium rct Full text arxiv
(Human) Attention Is (Still) All You Need: Human oversight makes AI-assisted social science reliable · Chen Zhu, Xiaolu Wang, Weilong Zhang · Jun 11, 2026
Accountants in Isabela use AI largely for routine tasks, improving efficiency and accuracy, but advanced applications are scarce; weak workforce skills and limited infrastructure are the main barriers to broader adoption, pointing to a need for targeted training and investment.
low descriptive Full text openalex
Utilization of Artificial Intelligence Technology among Accounting Firms in Isabela, Cagayan Valley: Towards Operational Efficiency · Rhodilet Batarao-Valdez · Jun 11, 2026
Emotional AI can lift engagement and short-term productivity in controlled studies, yet patchy evidence and risks of privacy violations, algorithmic bias and eroded trust limit confidence in broad deployments.
medium review_meta Full text openalex
Emotional AI in the Workplace: Systematic Review of Effects on Employee Well-Being, Productivity, and Organizational Performance · Jorge A. Ruíz-Vanoye, Jaime Aguilar-Ortíz, Ocotlán Diaz-Parra, Ricardo A. Barrer... · Jun 11, 2026