The Commonplace
Home Dashboard Papers Evidence Syntheses Digests 🎲

Papers (1134)

Filter: Sort:
A multi-agent VCA that feeds device diagnostics to LLMs more than doubles successful cybersecurity fixes, lifting correct resolutions from ~50% to over 90% and making users far more willing to replace human IT support; stepwise guidance also raises user satisfaction and cuts perceived burden.
medium quasi_experimental arxiv
SecMate: Multi-Agent Adaptive Cybersecurity Troubleshooting with Tri-Context Personalization · Yair Meidan, Omri Haller, Yulia Moshan, Shahaf David, Dudu Mimran, Yuval Elovici... · Apr 29, 2026
A new end-to-end benchmark for enterprise document-AI finds hybrid retrieval narrowly bests BM25 (nDCG@5 0.92 vs 0.91) and both outperform dense embedding (0.83), but parsing and retrieval quality weakly predict final answer correctness and systems often omit relevant facts despite high stated factual accuracy (85.5% vs completeness 0.40).
medium descriptive arxiv
Benchmarking Complex Multimodal Document Processing Pipelines: A Unified Evaluation Framework for Enterprise AI · Saurabh K. Singh, Sachin Raj · Apr 29, 2026
AI shopping agents leak consumers’ willingness-to-pay: sellers recover buyers’ prices nearly one-for-one from agents’ natural-language profiles, creating pervasive preference leakage; prompt tweaks won’t solve it, so system-level design must trade personalization for privacy.
medium rct arxiv
When Agents Shop for You: Role Coherence in AI-Mediated Markets · Soogand Alavi, Salar Nozari · Apr 29, 2026
Autonomous LLM trading agents handled $20m in ETH trades with 99.9% settlement success, but reliability was engineered around the model: prompt compilation, typed controls, validation and execution guards — not the base model alone. Targeted harness changes slashed fabricated sell-rule failures from 57% to 3% and raised capital deployment in the affected cohort from 42.9% to 78%.
medium descriptive arxiv
Operating-Layer Controls for Onchain Language-Model Agents Under Real Capital · T. J. Barton, Chris Constantakis, Patti Hauseman, Annie Mous, Alaska Hoffman, Br... · Apr 28, 2026
A robust, MILP‑constrained RL controller more than doubles simulated net profit for an EV ride‑hailing fleet—$1.22M versus $0.58M–$0.70M for strong baselines—while strictly avoiding electrical feeder violations.
low descriptive arxiv
Semi-Markov Reinforcement Learning for City-Scale EV Ride-Hailing with Feasibility-Guaranteed Actions · An Nguyen, Hoang Nguyen, Phuong Le, Hung Pham, Cuong Do, Laurent El Ghaoui · Apr 28, 2026
A flaw in standard bubble tests makes technology-driven price booms look speculative; after projecting out observable measures of technology adoption, the 2020–2025 AI rally shows no evidence of a speculative bubble while the dot‑com peak remains a genuine speculative episode.
medium theoretical arxiv
General-Purpose Technology and Speculative Bubble Detection · Haiqiang Chen, Li Chen, Difang Huang, Yuexin Li, Zhengjun Zhang · Apr 28, 2026
Salesforce's modular, serverless inference platform halved tail latency and cut inference costs by up to 40% while nearly quadrupling throughput for agentic enterprise AI; the design specifically mitigates compound-system issues such as multi-model fan-out and cascading cold-start propagation.
medium descriptive arxiv
Scalable Inference Architectures for Compound AI Systems: A Production Deployment Study · Srikanta Prasad S, Utkarsh Arora · Apr 28, 2026
Embedding governance inside LLM agents yields high compliance in one retail deployment: a four-layer Pre-Action Governance Reasoning Loop delivered 95% compliance and no false human escalations in a production supply-chain workflow; however, the result rests on a single proprietary implementation with limited methodological transparency.
low theoretical arxiv
Think Before You Act -- A Neurocognitive Governance Model for Autonomous AI Agents · Eranga Bandara, Ross Gore, Asanga Gunaratna, Sachini Rajapakse, Isurunima Kulara... · Apr 28, 2026
A semantic-search system indexed 166 million pediatric notes and cut clinician chart-abstraction time by a quarter to nearly nine‑tenths while serving queries in under a second at roughly $4,000/month; the deployment demonstrates that health-system‑scale semantic retrieval is technically and operationally feasible but evidence on broader clinical and cross‑site impacts remains limited.
medium quasi_experimental arxiv
Health System Scale Semantic Search Across Unstructured Clinical Notes · Faith Wavinya Mutinda, Spandana Makeneni, Anna Lin, Shivaji Dutta, Irit R. Rasoo... · Apr 28, 2026
AI mostly amplifies human workers rather than replacing them, but productivity payoffs depend on people: without widespread AI literacy and stronger complementary skills, technological gains concentrate among a few and create productivity bottlenecks; broad investment in workforce capabilities is therefore essential.
n/a review_meta semantic_scholar
AI as Augmentation: How Human Capital Shapes Technology's Impact on Productivity and Inequality · Jonathan H. Westover · Fetched Apr 28, 2026
Language models negotiate more reliably than humans in a simulated multi-player game and accept offers more frequently, and simple prompt-based behavior tweaks lift agent win rates from 22% to 33%.
medium descriptive arxiv
Cooperate to Compete: Strategic Coordination in Multi-Agent Conquest · Abigail O'Neill, Alan Zhu, Mihran Miroyan, Narges Norouzi, Joseph E. Gonzalez · Apr 28, 2026
An LLM-driven 'Agentic Architect' framework autonomously evolves microarchitecture components that match or beat state-of-the-art in cycle-accurate simulation — e.g., an evolved prefetcher yields a 1.76x geomean IPC boost over no prefetching and modest gains over top learned baselines; however, improvements are demonstrated in simulation and are contingent on seed quality and prompt guidance.
medium descriptive arxiv
Agentic Architect: An Agentic AI Framework for Architecture Design Exploration and Optimization · Alexander Blasberg, Vasilis Kypriotis, Dimitrios Skarlatos · Apr 28, 2026
Fear of AI is reshaping welfare politics: across OECD publics, anxiety about automation fuels support for measures that protect jobs from machines—such as robot taxes—more than for expanded unemployment benefits or retraining programs.
low correlational openalex
AI, the Future of Work, and the Politics of the Welfare State · Juliana Chueri · Apr 28, 2026
Generative AI is changing how firms coordinate by removing cognitive bottlenecks, prompting a new 'unbounded cognitive fusion' model in which coordination emerges from shared cognitive synthesis rather than prices or authority. The paper identifies three new organizational forms—cognitive meshworks, algorithmic ecosystems, and hybrid intelligence collectives—while noting limits where accountability, embodied skills, or low AI adoption persist.
n/a theoretical openalex
Beyond markets and hierarchies: How GenAI enables unbounded cognitive fusion · Yun Wan · Apr 28, 2026
LLMs reading headlines can build better-than-naive equity portfolios but don’t yet match AI-driven optimizers; low-turnover LLM strategies remain competitive after accounting for trading costs.
medium correlational openalex
Few-Shot Portfolio Optimization: Can Large Language Models Outperform Quantitative Portfolio Optimization? A Comparative Study of LLMs and Optimized Portfolio Allocators · Lamukanyani Alson Mantshimuli, John Weirstrass Muteba Mwamba · Apr 28, 2026
Fluent users drive AI to do harder, more ambitious work by iterating and critiquing outputs — they fail more often but these failures are visible and recoverable, yielding higher success on complex tasks, while novices more frequently encounter undetected, silent failures.
medium correlational openalex
A paradox of AI fluency · Christopher Potts, Moritz Sudhof · Apr 28, 2026
A formal leverage ratio shows AI agents can only replace human labor up to hard bounds set by information-transfer limits and irreducible planning, while repeated tasks can yield greater displacement only to the extent of prior planning investment.
n/a theoretical arxiv
Leverage Laws: A Per-Task Framework for Human-Agent Collaboration · Stan Loosmore · Apr 27, 2026
Chinese listed firms mirror their industry when adopting AI: average peer adoption drives firms’ own AI uptake more than a single leading adopter, though CIOs and high-market-position firms respond especially strongly to a peer leader.
medium correlational semantic_scholar
Following the Herd or the Bellwether: Peer Effects in Firms’ AI Adoption · Siyu Shao, Jianjun Yang, Ling Zhang · Fetched Apr 27, 2026
Large language models can act as strategic bidders in 6G spectrum auctions, replicating incentive-compatible outcomes when conditions permit and outperforming simple strategies under realistic budget constraints. AI bidders therefore have the potential to reshape auction dynamics and participant utilities in future wireless markets.
medium descriptive arxiv
Strategic Bidding in 6G Spectrum Auctions with Large Language Models · Ismail Lotfi, Ali Ghrayeb · Apr 27, 2026
A principled probability cutoff can stop exploitative Pascal-style gambles without breaking rational decision-making, offering implementable design norms for safer AI agents; practical calibration and deployment remain context-dependent and untested.
n/a theoretical semantic_scholar
Bounding the Long Tail: Ai Norms for Decision-Making Under Negligible Probabilities · Vipul Razdan · Fetched Apr 27, 2026
A 'Right-to-Act' protocol would place a deterministic veto at the point of AI execution, stopping any action if required conditions are unmet rather than letting high-confidence signals override failures; the approach promises safer reversibility but remains untested at scale and may impose practical trade-offs.
n/a theoretical arxiv
Right-to-Act: A Pre-Execution Non-Compensatory Decision Protocol for AI Systems · Gadi Lavi · Apr 27, 2026
Jointly training production and transport agents boosts scheduling performance in many simulated factories, but offers little benefit when severe transport or processing bottlenecks make one task dominant. In those bottlenecked settings, cheaper modular training or rule-based approaches can be a practical alternative.
medium descriptive arxiv
An Analysis of the Coordination Gap between Joint and Modular Learning for Job Shop Scheduling with Transportation Resources · Moritz Link, Jonathan Hoss, Noah Klarmann · Apr 27, 2026
Multi-agent tutoring systems are fast and affordable when hosted on priority pay-per-call plans, but standard paygo slows under classroom loads and reserved capacity only wins when utilization is reliably high; institutions should pick hosting tiers by expected concurrency to balance latency and cost.
medium descriptive arxiv
Latency and Cost of Multi-Agent Intelligent Tutoring at Scale · Iizalaarab Elhaimeur, Nikos Chrisochoides · Apr 27, 2026
AgentPulse: a continuous scoring system shows public adoption and community signals reveal deployment-relevant information that static benchmarks miss; those signals predict GitHub stars, Stack Overflow activity and IDE installs even where capability benchmarks do not.
medium descriptive arxiv
AgentPulse: A Continuous Multi-Signal Framework for Evaluating AI Agents in Deployment · Yuxuan Gao, Megan Wang, Yi Ling Yu · Apr 27, 2026
LLM coding assistants speed up developers and cut routine work, but their effect on code quality and teamwork remains unresolved; most studies are short-term and exploratory, leaving long-run and team-level impacts unclear.
medium review_meta openalex
The Impact of LLM-Assistants on Software Developer Productivity: A Systematic Review and Mapping Study · Amr Mohamed, Maram Assi, Mariam Guizani · Apr 27, 2026