Apple’s multi-year agreement to integrate Google’s Gemini models into its revamped Siri offers a rare window into how one of the world’s most selective technology companies evaluates foundation models – and the criteria should matter to any enterprise weighing similar decisions.

The stakes were considerable. Apple had been publicly integrating ChatGPT into its devices since late 2024, giving OpenAI prominent positioning in the Apple Intelligence ecosystem.

Google’s Gemini win represents a shift in Apple’s AI infrastructure strategy, one that relegates OpenAI to what Parth Talsania, CEO of Equisights Research, describes as “a more supporting role, with ChatGPT remaining positioned for complex, opt-in queries rather than the default intelligence layer.”

The evaluation that mattered

Apple’s reasoning was notably specific. “After careful evaluation, Apple determined Google’s AI technology provides the most capable foundation for Apple Foundation Models,” according to the joint statement. The phrasing matters – Apple didn’t cite partnership convenience, pricing, or ecosystem compatibility. The company framed this explicitly as a capabilities assessment.

Apple’s evaluation criteria likely mirrored concerns familiar to any organisation building AI into core products: model performance at scale, inference latency, multimodal capabilities, and crucially, the ability to run models both on-device and in cloud environments while maintaining privacy standards.

Google’s technology already powers Samsung’s Galaxy AI in millions of devices, providing proven deployment evidence at consumer scale. But Apple’s decision unlocks something different: integration in more than two billion active devices, with the technical demands that come with Apple’s performance and privacy requirements.

What has changed since ChatGPT integration

The timing raises questions. Apple rolled out ChatGPT integration just over a year ago, positioning Siri to tap into the chatbot for complex queries. The company now states, “there were no major changes to the ChatGPT integration at the time,” but the competitive dynamics have clearly shifted.

OpenAI’s response to Google’s Gemini 3 release in late 2025 – what reports described as a “code red” to accelerate development – suggests the competitive pressure was real. For enterprises, this highlights a risk often under-weighted in vendor selection: the pace of model capability advancement varies significantly between providers, and today’s leader may not maintain that position in a multi-year deployment.

Apple’s choice of a multi-year agreement with Google, rather than maintaining flexibility to switch between providers, suggests confidence in Google’s development trajectory. That’s a bet on sustained R&D investment, continued model improvements, and infrastructure scaling – the same factors enterprise buyers need to assess beyond current benchmarks.

The infrastructure question

The deal raises immediate concerns about concentration. “The seems like an unreasonable concentration of power for Google, given that they also have Android and Chrome,” Tesla CEO Elon Musk posted on social media. The critique reflects a legitimate enterprise concern about vendor dependency.

Google now powers AI features in both major mobile operating systems through different mechanisms: directly via Android, and through this partnership for iOS. For enterprises deploying AI capabilities, the parallel is that relying on a single foundation model provider creates technical and commercial dependencies that extend beyond the immediate integration.

This makes Apple’s architectural approach worth examining. The company emphasised that “Apple Intelligence will continue to run on Apple devices and Private Cloud Compute, while maintaining Apple’s industry-leading privacy standards.”

The hybrid deployment model – on-device processing for privacy-sensitive operations, cloud-based models for complex tasks – offers a template for enterprises balancing capability with data governance requirements.

Market implications beyond mobile

The deal’s immediate impact was measurable: Alphabet’s market valuation crossed US$4 trillion on Monday, with the stock having jumped 65% in 2024 on growing investor confidence in its AI efforts. But the strategic implications extend beyond market caps.

Google has been methodically building positions in the AI stack – frontier models, image and video generation, and now default integration into iOS devices. For enterprises, this vertical integration matters when evaluating cloud AI services: a provider’s foundation model capabilities increasingly connect to their broader infrastructure, tools, and ecosystem positioning.

Apple’s setbacks on the AI front – delayed Siri upgrades, executive changes, lukewarm reception for initial generative AI tools – are instructive from another angle. Even companies with enormous resources and talent can struggle with AI product execution. The decision to partner with Google rather than persist with entirely proprietary development acknowledges the complexity and resource demands of frontier model development.

The search revenue connection

The Gemini deal builds on an existing commercial relationship that generates tens of billions in annual revenue for Apple: Google pays to remain the default search engine on Apple devices. That arrangement has faced regulatory scrutiny, but it establishes precedent for deep technical integration between the companies.

The search deal likely influenced negotiations around the Gemini integration, just as existing vendor relationships shape enterprise AI procurement. Those relationships can be advantages – established trust, proven integration capabilities – or constraints that limit evaluation of alternatives.

The OpenAI question

The deal leaves OpenAI in an awkward position. ChatGPT remains available on Apple devices, but as an optional feature rather than the infrastructure layer. For a company that has positioned itself as the AI leader, losing default integration to Google represents a strategic setback.

The competitive dynamic offers a reminder that the foundation model market remains fluid. Provider positioning can shift quickly, and exclusive relationships between major players can reshape options for everyone else. Maintaining options – through abstraction layers, multi-model strategies, or portable architectures – becomes more valuable in rapidly evolving markets.

What comes next

Google stated that Gemini models will power not just the revamped Siri coming later this year, but “other future Apple Intelligence features.” The scope of integration will likely expand as Apple builds out its AI capabilities, creating deeper technical dependencies and raising the stakes of the partnership.

The financial terms remain undisclosed, leaving the question of how Apple and Google structure pricing for this scale of deployment? Enterprise buyers negotiating foundation model licensing will be watching for any signals about how such deals get priced at a massive scale.

Apple’s decision doesn’t make Google’s Gemini the obvious choice for every enterprise – far from it. But the deal does offer validated evidence of what one extremely selective technology company prioritised when evaluating foundation models under demanding requirements. For enterprise AI buyers navigating their own evaluations, that’s a signal worth considering amid the noise of vendor marketing and benchmark leader boards.

See also: Apple plans big Siri update with help from Google AI

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and co-located with other leading technology events. Click here for more information.

The post Why Apple chose Google over OpenAI: What enterprise AI buyers can learn from the Gemini deal appeared first on AI News.

arXiv:2601.06161v1 Announce Type: new Abstract: Artificial intelligence (AI) systems increasingly achieve expert-level predictive accuracy in healthcare, yet improvements in model performance often fail to produce corresponding gains in patient outcomes. We term this disconnect the allocation gap and provide a decision-theoretic explanation by modelling healthcare delivery as a stochastic allocation problem under binding resource constraints. In this framework, AI acts as decision infrastructure that estimates utility rather than making autonomous decisions. Using constrained optimisation and Markov decision processes, we show how improved estimation affects optimal allocation under scarcity. A synthetic triage simulation demonstrates that allocation-aware policies substantially outperform risk-threshold approaches in realised utility, even with identical predictive accuracy. The framework provides a principled basis for evaluating and deploying healthcare AI in resource-constrained settings.

arXiv:2601.06197v1 Announce Type: new Abstract: Generative AI has unleashed the power of content generation and it has also unwittingly opened the pandora box of realistic deepfake causing a number of social hazards and harm to businesses and personal reputation. The investigation & ramification of Generative AI technology across industries, the resolution & hybridization detection techniques using neural networks allows flagging of the content. Good detection techniques & flagging allow AI safety - this is the main focus of this paper. The research provides a significant method for efficiently detecting dark side problems by imposing a Temporal Consistency Learning (TCL) technique. Through pretrained Temporal Convolutional Networks (TCNs) model training and performance comparison, this paper showcases that TCN models outperforms the other approaches and achieves significant accuracy for five dark side problems. Findings highlight how important it is to take proactive measures in identification to reduce any potential risks associated with generative artificial intelligence.

arXiv:2601.06453v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly grounded in sensor data to perceive and reason about human physiology and the physical world. However, accurately interpreting heterogeneous multimodal sensor data remains a fundamental challenge. We show that a single monolithic LLM often fails to reason coherently across modalities, leading to incomplete interpretations and prior-knowledge bias. We introduce ConSensus, a training-free multi-agent collaboration framework that decomposes multimodal sensing tasks into specialized, modality-aware agents. To aggregate agent-level interpretations, we propose a hybrid fusion mechanism that balances semantic aggregation, which enables cross-modal reasoning and contextual understanding, with statistical consensus, which provides robustness through agreement across modalities. While each approach has complementary failure modes, their combination enables reliable inference under sensor noise and missing data. We evaluate ConSensus on five diverse multimodal sensing benchmarks, demonstrating an average accuracy improvement of 7.1% over the single-agent baseline. Furthermore, ConSensus matches or exceeds the performance of iterative multi-agent debate methods while achieving a 12.7 times reduction in average fusion token cost through a single-round hybrid fusion protocol, yielding a robust and efficient solution for real-world multimodal sensing tasks.

arXiv:2601.06500v1 Announce Type: new Abstract: Artificial intelligence (AI) represents a qualitative shift in technological change by extending cognitive labor itself rather than merely automating routine tasks. Recent evidence shows that generative AI disproportionately affects highly educated, white collar work, challenging existing assumptions about workforce vulnerability and rendering traditional approaches to digital or AI literacy insufficient. This paper introduces the concept of AI Nativity, the capacity to integrate AI fluidly into everyday reasoning, problem solving, and decision making, and proposes the AI Pyramid, a conceptual framework for organizing human capability in an AI mediated economy. The framework distinguishes three interdependent capability layers: AI Native capability as a universal baseline for participation in AI augmented environments; AI Foundation capability for building, integrating, and sustaining AI enabled systems; and AI Deep capability for advancing frontier AI knowledge and applications. Crucially, the pyramid is not a career ladder but a system level distribution of capabilities required at scale. Building on this structure, the paper argues that effective AI workforce development requires treating capability formation as infrastructure rather than episodic training, centered on problem based learning embedded in work contexts and supported by dynamic skill ontologies and competency based measurement. The framework has implications for organizations, education systems, and governments seeking to align learning, measurement, and policy with the evolving demands of AI mediated work, while addressing productivity, resilience, and inequality at societal scale.

arXiv:2601.06663v1 Announce Type: new Abstract: Large language model-based agents are rapidly evolving from simple conversational assistants into autonomous systems capable of performing complex, professional-level tasks in various domains. While these advancements promise significant productivity gains, they also introduce critical safety risks that remain under-explored. Existing safety evaluations primarily focus on simple, daily assistance tasks, failing to capture the intricate decision-making processes and potential consequences of misaligned behaviors in professional settings. To address this gap, we introduce \textbf{SafePro}, a comprehensive benchmark designed to evaluate the safety alignment of AI agents performing professional activities. SafePro features a dataset of high-complexity tasks across diverse professional domains with safety risks, developed through a rigorous iterative creation and review process. Our evaluation of state-of-the-art AI models reveals significant safety vulnerabilities and uncovers new unsafe behaviors in professional contexts. We further show that these models exhibit both insufficient safety judgment and weak safety alignment when executing complex professional tasks. In addition, we investigate safety mitigation strategies for improving agent safety in these scenarios and observe encouraging improvements. Together, our findings highlight the urgent need for robust safety mechanisms tailored to the next generation of professional AI agents.

arXiv:2601.07233v1 Announce Type: new Abstract: Explainable AI (XAI) in high-stakes domains should help stakeholders trust and verify system outputs. Yet Chain-of-Thought methods reason before concluding, and logical gaps or hallucinations can yield conclusions that do not reliably align with their rationale. Thus, we propose "Result -> Justify", which constrains the output communication to present a conclusion before its structured justification. We introduce SEF (Structured Explainability Framework), operationalizing professional conventions (e.g., CREAC, BLUF) via six metrics for structure and grounding. Experiments across four tasks in three domains validate this approach: all six metrics correlate with correctness (r=0.20-0.42; p

arXiv:2601.07463v1 Announce Type: new Abstract: Offline multi-agent reinforcement learning (MARL) aims to solve cooperative decision-making problems in multi-agent systems using pre-collected datasets. Existing offline MARL methods primarily constrain training within the dataset distribution, resulting in overly conservative policies that struggle to generalize beyond the support of the data. While model-based approaches offer a promising solution by expanding the original dataset with synthetic data generated from a learned world model, the high dimensionality, non-stationarity, and complexity of multi-agent systems make it challenging to accurately estimate the transitions and reward functions in offline MARL. Given the difficulty of directly modeling joint dynamics, we propose a local-to-global (LOGO) world model, a novel framework that leverages local predictions-which are easier to estimate-to infer global state dynamics, thus improving prediction accuracy while implicitly capturing agent-wise dependencies. Using the trained world model, we generate synthetic data to augment the original dataset, expanding the effective state-action space. To ensure reliable policy learning, we further introduce an uncertainty-aware sampling mechanism that adaptively weights synthetic data by prediction uncertainty, reducing approximation error propagation to policies. In contrast to conventional ensemble-based methods, our approach requires only an additional encoder for uncertainty estimation, significantly reducing computational overhead while maintaining accuracy. Extensive experiments across 8 scenarios against 8 baselines demonstrate that our method surpasses state-of-the-art baselines on standard offline MARL benchmarks, establishing a new model-based baseline for generalizable offline multi-agent learning.

arXiv:2601.07611v1 Announce Type: new Abstract: Paper weakness identification using single-agent or multi-agent LLMs has attracted increasing attention, yet existing approaches exhibit key limitations. Many multi-agent systems simulate human roles at a surface level, missing the underlying criteria that lead experts to assess complementary intellectual aspects of a paper. Moreover, prior methods implicitly assume identified weaknesses are valid, ignoring reviewer bias, misunderstanding, and the critical role of author rebuttals in validating review quality. Finally, most systems output unranked weakness lists, rather than prioritizing the most consequential issues for users. In this work, we propose DIAGPaper, a novel multi-agent framework that addresses these challenges through three tightly integrated modules. The customizer module simulates human-defined review criteria and instantiates multiple reviewer agents with criterion-specific expertise. The rebuttal module introduces author agents that engage in structured debate with reviewer agents to validate and refine proposed weaknesses. The prioritizer module learns from large-scale human review practices to assess the severity of validated weaknesses and surfaces the top-K severest ones to users. Experiments on two benchmarks, AAAR and ReviewCritique, demonstrate that DIAGPaper substantially outperforms existing methods by producing more valid and more paper-specific weaknesses, while presenting them in a user-oriented, prioritized manner.

arXiv:2601.06030v1 Announce Type: cross Abstract: This paper offers a concise, 60-year synthesis of human-AI collaboration, from Licklider's ``man-computer symbiosis" (AI as colleague) and Engelbart's ``augmenting human intellect" (AI as tool) to contemporary poles: Human-Centered AI's ``supertool" and Symbiotic Intelligence's mutual-adaptation model. We formalize the mechanism for effective teaming as a causal chain: Explainable AI (XAI) -> co-adaptation -> shared mental models (SMMs). A meta-analytic ``performance paradox" is then examined: human-AI teams tend to show negative synergy in judgment/decision tasks (underperforming AI alone) but positive synergy in content creation and problem formulation. We trace failures to the algorithm-in-the-loop dynamic, aversion/bias asymmetries, and cumulative cognitive deskilling. We conclude with a unifying framework--combining extended-self and dual-process theories--arguing that durable gains arise when AI functions as an internalized cognitive component, yielding a unitary human-XAI symbiotic agency. This resolves the paradox and delineates a forward agenda for research and practice.

arXiv:2601.06031v1 Announce Type: cross Abstract: Graphical user interface (GUI) grounding, the process of mapping human instructions to GUI actions, serves as a fundamental basis to autonomous GUI agents. While existing grounding models achieve promising performance to simulate the mouse click action on various click-based benchmarks, another essential mode of mouse interaction, namely dragging, remains largely underexplored. Yet, dragging the mouse to select and manipulate textual content represents a prevalent and important usage in practical GUI scenarios. To narrow this gap, we first introduce GUI-Drag, a diverse dataset of 161K text dragging examples synthesized through a scalable pipeline. To support systematic and robust evaluation, we further construct ScreenDrag, a benchmark with 5,333 examples spanning three levels of interface context, together with three dedicated metrics designed for assessing text dragging capability. Models trained on GUI-Drag with an efficient continual training strategy achieve substantial improvements on ScreenDrag, while preserving the original click-based performance on ScreenSpot, ScreenSpot-v2, and OSWorld-G. Our work encourages further research on broader GUI grounding beyond just clicking and paves way toward a truly generalist GUI grounding model. All benchmark, data, checkpoints, and code are open-sourced and available at https://osu-nlp-group.github.io/GUI-Drag.

arXiv:2601.06048v1 Announce Type: cross Abstract: This paper examines the admissibility of AI-generated forensic evidence in criminal trials. The growing adoption of AI presents promising results for investigative efficiency. Despite advancements, significant research gaps persist in practically understanding the legal limits of AI evidence in judicial processes. Existing literature lacks focused assessment of the evidentiary value of AI outputs. The objective of this study is to evaluate whether AI-generated evidence satisfies established legal standards of reliability. The methodology involves a comparative doctrinal legal analysis of evidentiary standards across common law jurisdictions. Preliminary results indicate that AI forensic tools can enhance scale of evidence analysis. However, challenges arise from reproducibility deficits. Courts exhibit variability in acceptance of AI evidence due to limited technical literacy and lack of standardized validation protocols. Liability implications reveal that developers and investigators may bear accountability for flawed outputs. This raises critical concerns related to wrongful conviction. The paper emphasizes the necessity of independent validation and, development of AI-specific admissibility criteria. Findings inform policy development for the responsible AI integration within criminal justice systems. The research advances the objectives of Sustainable Development Goal 16 by reinforcing equitable access to justice. Preliminary results contribute for a foundation for future empirical research in AI deployed criminal forensics.

arXiv:2601.06060v1 Announce Type: cross Abstract: AI-generated "slop" is often seen as digital pollution. We argue that this dismissal of the topic risks missing important aspects of AI Slop that deserve rigorous study. AI Slop serves a social function: it offers a supply-side solution to a variety of problems in cultural and economic demand - that, collectively, people want more content than humans can supply. We also argue that AI Slop is not mere digital detritus but has its own aesthetic value. Like other "low" cultural forms initially dismissed by critics, it nonetheless offers a legitimate means of collective sense-making, with the potential to express meaning and identity. We identify three key features of family resemblance for prototypical AI Slop: superficial competence (its veneer of quality is belied by a deeper lack of substance), asymmetry effort (it takes vastly less effort to generate than would be the case without AI), and mass producibility (it is part of a digital ecosystem of widespread generation and consumption). While AI Slop is heterogeneous and depends crucially on its medium, it tends to vary across three dimensions: instrumental utility, personalization, and surrealism. AI Slop will be an increasingly prolific and impactful part of our creative, information, and cultural economies; we should take it seriously as an object of study in its own right.

arXiv:2601.06144v1 Announce Type: cross Abstract: Artificial intelligence (AI) in healthcare has led to many promising developments; however, increasingly, AI research is funded by the private sector leading to potential trade-offs between benefits to patients and benefits to industry. Health AI practitioners should prioritize successful adaptation into clinical practice in order to provide meaningful benefits to patients, but translation usually requires collaboration with industry. We discuss three features of AI studies that hamper the integration of AI into clinical practice from the perspective of researchers and clinicians. These include lack of clinically relevant metrics, lack of clinical trials and longitudinal studies to validate results, and lack of patient and physician involvement in the development process. For partnerships between industry and health research to be sustainable, a balance must be established between patient and industry benefit. We propose three approaches for addressing this gap: improved transparency and explainability of AI models, fostering relationships with industry partners that have a reputation for centering patient benefit in their practices, and prioritization of overall healthcare benefits. With these priorities, we can sooner realize meaningful AI technologies used by clinicians where mutua

arXiv:2601.06153v1 Announce Type: cross Abstract: This policy report draws on country studies from China, South Korea, Singapore, and the United Kingdom to identify effective tools and key barriers to interoperability in AI safety governance. It offers practical recommendations to support a globally informed yet locally grounded governance ecosystem. Interoperability is a central goal of AI governance, vital for reducing risks, fostering innovation, enhancing competitiveness, promoting standardization, and building public trust. However, structural gaps such as fragmented regulations and lack of global coordination, and conceptual gaps, including limited Global South engagement, continue to hinder progress. Focusing on three high-stakes domains - autonomous vehicles, education, and cross-border data flows - the report compares ethical, legal, and technical frameworks across the four countries. It identifies areas of convergence, divergence, and potential alignment, offering policy recommendations that support the development of interoperability mechanisms aligned with the Global Digital Compact and relevant UN resolutions. The analysis covers seven components: objectives, regulators, ethics, binding measures, targeted frameworks, technical standards, and key risks.

arXiv:2601.06223v1 Announce Type: cross Abstract: This paper presents a conceptual and operational framework for developing and operating safe and trustworthy AI agents based on a Three-Pillar Model grounded in transparency, accountability, and trustworthiness. Building on prior work in Human-in-the-Loop systems, reinforcement learning, and collaborative AI, the framework defines an evolutionary path toward autonomous agents that balances increasing automation with appropriate human oversight. The paper argues that safe agent autonomy must be achieved through progressive validation, analogous to the staged development of autonomous driving, rather than through immediate full automation. Transparency and accountability are identified as foundational requirements for establishing user trust and for mitigating known risks in generative AI systems, including hallucinations, data bias, and goal misalignment, such as the inversion problem. The paper further describes three ongoing work streams supporting this framework: public deliberation on AI agents conducted by the Stanford Deliberative Democracy Lab, cross-industry collaboration through the Safe AI Agent Consortium, and the development of open tooling for an agent operating environment aligned with the Three-Pillar Model. Together, these contributions provide both conceptual clarity and practical guidance for enabling the responsible evolution of AI agents that operate transparently, remain aligned with human values, and sustain societal trust.

arXiv:2601.06300v1 Announce Type: cross Abstract: Clinical trial amendments frequently introduce delays, increased costs, and administrative burden, with eligibility criteria being the most commonly amended component. We introduce \textit{eligibility criteria amendment prediction}, a novel NLP task that aims to forecast whether the eligibility criteria of an initial trial protocol will undergo future amendments. To support this task, we release $\texttt{AMEND++}$, a benchmark suite comprising two datasets: $\texttt{AMEND}$, which captures eligibility-criteria version histories and amendment labels from public clinical trials, and $\verb|AMEND_LLM|$, a refined subset curated using an LLM-based denoising pipeline to isolate substantive changes. We further propose $\textit{Change-Aware Masked Language Modeling}$ (CAMLM), a revision-aware pretraining strategy that leverages historical edits to learn amendment-sensitive representations. Experiments across diverse baselines show that CAMLM consistently improves amendment prediction, enabling more robust and cost-effective clinical trial design.

arXiv:2601.06364v1 Announce Type: cross Abstract: Chronic disease management requires regular adherence feedback to prevent avoidable hospitalizations, yet clinicians lack time to produce personalized patient communications. Manual authoring preserves clinical accuracy but does not scale; AI generation scales but can undermine trust in patient-facing contexts. We present a clinician-in-the-loop interface that constrains AI to data organization and preserves physician oversight through recognition-based review. A single-page editor pairs AI-generated section drafts with time-aligned visualizations, enabling inline editing with visual evidence for each claim. This division of labor (AI organizes, clinician decides) targets both efficiency and accountability. In a pilot with three physicians reviewing 24 cases, AI successfully generated clinically personalized drafts matching physicians' manual authoring practice (overall mean 4.86/10 vs. 5.0/10 baseline), requiring minimal physician editing (mean 8.3\% content modification) with zero safety-critical issues, demonstrating effective automation of content generation. However, review time remained comparable to manual practice, revealing an accountability paradox: in high-stakes clinical contexts, professional responsibility requires complete verification regardless of AI accuracy. We contribute three interaction patterns for clinical AI collaboration: bounded generation with recognition-based review via chart-text pairing, automated urgency flagging that analyzes vital trends and adherence patterns with fail-safe escalation for missed critical monitoring tasks, and progressive disclosure controls that reduce cognitive load while maintaining oversight. These patterns indicate that clinical AI efficiency requires not only accurate models, but also mechanisms for selective verification that preserve accountability.

arXiv:2601.06750v1 Announce Type: cross Abstract: Medical Multimodal Large Language Models (Med-MLLMs) require egocentric clinical intent understanding for real-world deployment, yet existing benchmarks fail to evaluate this critical capability. To address these challenges, we introduce MedGaze-Bench, the first benchmark leveraging clinician gaze as a Cognitive Cursor to assess intent understanding across surgery, emergency simulation, and diagnostic interpretation. Our benchmark addresses three fundamental challenges: visual homogeneity of anatomical structures, strict temporal-causal dependencies in clinical workflows, and implicit adherence to safety protocols. We propose a Three-Dimensional Clinical Intent Framework evaluating: (1) Spatial Intent: discriminating precise targets amid visual noise, (2) Temporal Intent: inferring causal rationale through retrospective and prospective reasoning, and (3) Standard Intent: verifying protocol compliance through safety checks. Beyond accuracy metrics, we introduce Trap QA mechanisms to stress-test clinical reliability by penalizing hallucinations and cognitive sycophancy. Experiments reveal current MLLMs struggle with egocentric intent due to over-reliance on global features, leading to fabricated observations and uncritical acceptance of invalid instructions.

arXiv:2601.07136v1 Announce Type: cross Abstract: The rapid emergence of multi-agent AI systems (MAS), including LangChain, CrewAI, and AutoGen, has shaped how large language model (LLM) applications are developed and orchestrated. However, little is known about how these systems evolve and are maintained in practice. This paper presents the first large-scale empirical study of open-source MAS, analyzing over 42K unique commits and over 4.7K resolved issues across eight leading systems. Our analysis identifies three distinct development profiles: sustained, steady, and burst-driven. These profiles reflect substantial variation in ecosystem maturity. Perfective commits constitute 40.8% of all changes, suggesting that feature enhancement is prioritized over corrective maintenance (27.4%) and adaptive updates (24.3%). Data about issues shows that the most frequent concerns involve bugs (22%), infrastructure (14%), and agent coordination challenges (10%). Issue reporting also increased sharply across all frameworks starting in 2023. Median resolution times range from under one day to about two weeks, with distributions skewed toward fast responses but a minority of issues requiring extended attention. These results highlight both the momentum and the fragility of the current ecosystem, emphasizing the need for improved testing infrastructure, documentation quality, and maintenance practices to ensure long-term reliability and sustainability.