{"id":"318233b3-ab33-4dfd-a3c9-ca939f857fb7","arxiv_id":"2606.01015","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":3.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"A literature survey of AI-IoT-Robotics integration synthesizes pairwise work, identifies gaps in full three-way systems, and proposes a modular hybrid SLM-LLM architecture for connected robotics.","lead":"This survey reviews frameworks combining AI, IoT, and robotics and proposes a modular architecture using small language models at the edge and large ones in the cloud. A smart generalist might read it to understand trends toward connected robotic systems that sense, reason, and act in real time.","discovery_kind":"review","skeptic_critique":{"model":"grok-4.3","headline":"No significant objection identified","rationale":"The reader's assessment of the survey nature and the modular-architecture assumption is accurate. Because the work makes no falsifiable technical claim beyond literature synthesis, the identified assumption does not constitute an internal load-bearing flaw requiring verdict change.","tokens_in":1725,"tokens_out":245,"duration_ms":15051,"concrete_test":"Extract the survey's section classifying work by integration depth and list the specific cited references claimed to demonstrate simultaneous gains in adaptation, scalability, and reliability via hybrid SLM-LLM + IoT + robotics; confirm each reference actually reports all three metrics together.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper is a survey that synthesizes pairwise integrations (AIoT, IoRT) and proposes a conceptual modular architecture for full AI-IoT-robotics convergence using hybrid SLM-LLM systems. The central claim—that the review shows such hybrids can address real-time adaptation, scalability, and reliability—is presented as an emerging trend and roadmap rather than a verified result. No equations, proofs, new experiments, or quantitative benchmarks are introduced that would require independent validation beyond accurate citation of prior work.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper is a survey synthesizing pairwise integrations of AI with IoT (AIoT) and robotics (IoRT), proposing a conceptual modular architecture that combines Small Language Models at the edge with Large Language Models in the cloud for distributed cognition, classifying prior work by integration depth, identifying gaps in interoperability and feedback control, and presenting a roadmap toward the paradigms of Connected Robotics and Physical AI. The central claim is that hybrid SLM-LLM systems coupled with IoT and robotic agents can address real-time adaptation, scalability, and reliability challenges.","tokens_in":1813,"tokens_out":482,"duration_ms":17780,"significance":"If the synthesis is comprehensive and the classification scheme proves reproducible, the manuscript could provide a useful organizing framework and timely overview for the emerging intersection of these fields, highlighting the shift from pairwise to tripartite integration and the role of hybrid language models. The explicit roadmap and gap analysis are strengths for guiding future systems work in robotics and IoT.","major_comments":[{"comment":"The section classifying existing work by integration depth does not define or operationalize the criteria for assigning papers to depth categories (e.g., no taxonomy table, decision procedure, or inter-rater reliability discussion), which directly undermines the claim of systematic classification and makes it impossible to verify completeness or bias in the synthesis.","section":"Classification of Existing Work"},{"comment":"In the description of the proposed modular architecture, the manuscript asserts that the hybrid SLM-LLM design resolves gaps in feedback control and interoperability but provides no concrete interface specifications, data-flow diagrams, or comparison against existing pairwise systems that would demonstrate resolution without introducing new incompatibilities.","section":"Proposed Modular System Architecture"}],"minor_comments":[{"comment":"The abstract and introduction introduce the terms 'Connected Robotics' and 'Physical AI' as emerging paradigms without providing even a brief working definition or citation to prior usage, which reduces clarity for readers.","section":"Abstract and Introduction"},{"comment":"Several citations to recent SLM and LLM robotics papers appear without discussion of their specific limitations in real-time IoT settings, which would strengthen the gap analysis.","section":"State-of-the-Art Review"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive comments, which help improve the clarity and rigor of our survey. We provide point-by-point responses to the major comments below.","responses":[{"response":"We agree that the original manuscript did not provide explicit operational criteria for the integration depth classification. This is a valid point. In the revised manuscript, we will introduce a taxonomy table that clearly defines the categories (e.g., Level 1: Pairwise AI-IoT, Level 2: IoRT, Level 3: Full AI-IoT-Robotics) and a step-by-step decision procedure for classifying papers. Although inter-rater reliability metrics are uncommon in survey papers without multiple coders, the added procedure will allow readers to reproduce the classification.","revision_made":"yes","referee_comment":"[Classification of Existing Work] The section classifying existing work by integration depth does not define or operationalize the criteria for assigning papers to depth categories (e.g., no taxonomy table, decision procedure, or inter-rater reliability discussion), which directly undermines the claim of systematic classification and makes it impossible to verify completeness or bias in the synthesis."},{"response":"The architecture is presented as a high-level conceptual proposal aligned with emerging trends, rather than a detailed engineering specification. To strengthen this section, we will add data-flow diagrams illustrating the SLM-LLM interaction via IoT middleware, example interface specifications (such as standardized message formats for feedback loops), and a brief comparison table against selected pairwise systems (e.g., AIoT and IoRT) to show how the tripartite integration mitigates the gaps without introducing incompatibilities.","revision_made":"yes","referee_comment":"[Proposed Modular System Architecture] In the description of the proposed modular architecture, the manuscript asserts that the hybrid SLM-LLM design resolves gaps in feedback control and interoperability but provides no concrete interface specifications, data-flow diagrams, or comparison against existing pairwise systems that would demonstrate resolution without introducing new incompatibilities."}],"tokens_in":1384,"tokens_out":435,"duration_ms":21285,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main thing with this paper is that it's a survey synthesizing the state of AI, IoT, and robotics integration, with a proposal for a modular architecture using hybrid small and large language models. It doesn't present new experiments or proofs.\n\nWhat it does well is to bring out the limitations of the existing pairwise integrations and suggest how a three-way setup could help with scalability and adaptation in robotic systems. The classification of work by how deeply the components are integrated is a useful way to map the field, and the focus on edge processing with SLMs paired with cloud LLMs aligns with practical constraints in robotics.\n\nThe soft spots are around the lack of detail on how the literature was reviewed or classified. Without explicit criteria or a list of sources, it's tough to gauge if the identified gaps are comprehensive. The architecture is described at a high level, so it doesn't come with evidence that it would actually solve the interoperability issues better than current methods.\n\nThis is the kind of paper that would be helpful for someone trying to get up to speed on connected robotics ideas or for teams planning system designs across these domains. It won't be a core reference for specialists, but it could organize discussions.\n\nI would take it to a reading group if we're looking at system-level robotics papers. I probably wouldn't cite it myself unless I'm writing a similar survey. It deserves peer review because the synthesis could be refined with community input on the trends and the proposal.","headline":"This is a survey that organizes existing AI-IoT-robotics work by integration depth and sketches a modular hybrid SLM-LLM architecture, without new data or validated claims.","tokens_in":2293,"tokens_out":372,"would_cite":false,"duration_ms":21585,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"Hybrid SLM-LLM systems paired with IoT infrastructure and robotic agents address real-time adaptation, scalability, and reliability gaps.","keywords":["AI-IoT integration","robotics","small language models","large language models","modular architecture","connected robotics","interoperability","physical AI"],"falsifier":"An implemented three-way system that achieves full real-time adaptation, scalability, and reliability using only existing pairwise methods without any modular SLM-LLM layering, or a modular prototype that introduces new control incompatibilities not present in the pairwise cases.","tokens_in":2628,"feed_emoji":"🤖","tokens_out":678,"duration_ms":13493,"temperature":0.7,"pith_summary":"This survey reviews frameworks across AI, IoT, and robotics domains and finds that pairwise integrations like AIoT and IoRT have advanced but lack unified designs for all three together. It identifies persistent gaps in interoperability and feedback control, then proposes a modular architecture that places small language models at the edge for local tasks and large language models in the cloud for higher reasoning. The central claim is that coupling these hybrid models with IoT sensing and robotic actuation enables distributed cognition and autonomous decisions in changing environments. A sympathetic reader would care because the architecture offers a concrete roadmap for systems that adapt in real time without sacrificing scale or reliability.","feed_headline":"Hybrid edge-cloud models fix robotics integration gaps","feed_subtitle":"Survey shows modular SLM-LLM layering with IoT sensing closes interoperability and control shortfalls for real-time robotic systems.","key_machinery":"The modular system architecture that places SLMs at the edge and LLMs in the cloud to integrate AI perception, IoT communication, and robotic actuation for distributed cognition.","core_discovery":"The paper establishes that a modular system architecture aligning small language models at the edge with large language models in the cloud, combined with IoT sensing and robotic actuation, overcomes documented gaps in interoperability and feedback control. This setup supports distributed cognition and autonomous decision-making, allowing systems to handle real-time adaptation, scalability, and reliability challenges that current pairwise integrations leave unresolved.","pith_inferences":["Testing the proposed modular layers on a physical robot fleet could reveal whether edge SLMs reduce cloud latency enough to meet safety-critical timing needs.","Standardized interfaces between SLM outputs and IoT data streams might emerge as a practical next step the survey leaves open.","The same architecture could apply to non-robotic domains like smart infrastructure if the feedback control gaps prove domain-independent."],"forward_implications":["Existing work can be classified by integration depth to guide future designs.","Hybrid SLM-LLM coupling with IoT and robots delivers real-time adaptation in dynamic settings.","Scalability improves through distributed edge-cloud processing.","Reliability increases via better feedback control loops across the three domains.","The architecture provides a technical roadmap for connected robotics and physical AI ecosystems."],"fun_headline_variants":["Modular edge cloud models for AI IoT robotics","Survey maps SLM LLM use in connected robotics","Frameworks for AI IoT robotics convergence","Path to connected robotics via modular design","Trends in hybrid AI IoT robotic systems"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"Documented gaps in interoperability and feedback control between AI, IoT, and robotics can be closed by a modular architecture without creating new incompatibilities beyond those already seen in pairwise integrations.","fun_headline_variants_meta":{"raw":{"variants":["Modular edge cloud models for AI IoT robotics","Survey maps SLM LLM use in connected robotics","Frameworks for AI IoT robotics convergence","Path to connected robotics via modular design","Trends in hybrid AI IoT robotic systems"]},"model":"grok-4.3","cost_usd":0.007264,"raw_usage":{"total_tokens":3352,"prompt_tokens":678,"num_sources_used":0,"completion_tokens":65,"cost_in_usd_ticks":72637000,"prompt_tokens_details":{"text_tokens":678,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2609,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":678,"tokens_out":65,"duration_ms":24819,"temperature":1.0,"reasoning_tokens":2609,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-28T17:21:46.813049+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"An implemented three-way system that achieves full real-time adaptation, scalability, and reliability using only existing pairwise methods without any modular SLM-LLM layering, or a modular prototype that introduces new control incompatibilities not present in the pairwise cases.","supporting_citations":[],"review_version":1}