{"id":"f49b2c2a-eef3-4cdd-a2ef-9c061c0b30aa","arxiv_id":"2605.24662","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":2,"one_line_summary":"OpenTwin is a digital twin system for O-RAN that uses XGBoost and time-aware RLS to achieve up to 96% KPM fidelity and safe closed-loop control without live-network disruption.","lead":"The paper presents OpenTwin, a digital twin framework for Open RAN that combines an XGBoost model to set simulator parameters with a recursive least squares tuner to keep virtual KPMs aligned with real measurements. This setup lets developers test energy-saving control policies in simulation before deploying them on live networks.","discovery_kind":"unclear","skeptic_critique":{"model":"grok-4.3","headline":"Twin fidelity under traffic/config changes depends on uncharacterized duration of XGBoost+RLS correction before resync threshold breach","rationale":"The reader's weakest_assumption exactly isolates the load-bearing mechanism (XGBoost+RLS keeping deviation below threshold under change). Full-text availability does not alter this because the abstract already surfaces the assumption and the provided summary supplies no counter-evidence that would falsify it. Therefore the reader's UNVERDICTED stance remains appropriate.","tokens_in":1732,"tokens_out":337,"duration_ms":13989,"concrete_test":"From the energy-saving xApp experiment logs, extract the distribution of intervals between resynchronization triggers and the KPM deviation value immediately prior to each trigger; if median interval < 5 min or pre-trigger deviation > 8% on >20% of triggers under the reported traffic traces, the sustained-fidelity claim is unsupported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline claims (96% KPM mirroring accuracy; energy xApp reduces consumption without live disruption) require that the XGBoost configuration generator plus time-aware RLS tuner keep simulator-to-real deviation below the deviation-aware scoring threshold for operationally relevant intervals, even after network traffic or configuration shifts. The abstract states a deviation-aware mechanism triggers resynchronization on drift, yet provides no quantitative data on (a) typical time-to-drift, (b) deviation growth rate under realistic non-stationary traffic, or (c) whether resync itself introduces measurable control latency or service impact. This leaves the core closed-loop assumption untested at the timescale that would matter for continuous xApp operation.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper introduces OpenTwin, a digital twin framework for O-RAN built on ns-O-RAN-flexRIC and O1 KPM streaming in the non-RT RIC. It employs a two-step ML approach—an XGBoost model to learn and generate time-varying simulator configuration parameters, followed by a time-aware recursive least squares (RLS) tuner for continuous KPM deviation correction—together with a deviation-aware scoring mechanism that triggers resynchronization on detected drift. The framework is demonstrated via an energy-saving xApp that validates policies in the twin before live application, claiming up to 96% KPM mirroring accuracy and significant energy reduction without disrupting operations.","tokens_in":1866,"tokens_out":541,"duration_ms":19049,"significance":"If the twin fidelity can be sustained over operationally relevant intervals under traffic and configuration changes, OpenTwin would address key barriers to ML xApp/rApp development in O-RAN by enabling safe, data-efficient validation outside live networks. The combination of configuration generation and online correction is a pragmatic approach to simulator-to-reality alignment, though its practical value hinges on uncharacterized closed-loop duration.","major_comments":[{"comment":"Evaluation section (energy-saving xApp experiments): the headline claims of 96% KPM accuracy and disruption-free control require that the XGBoost+RLS loop keeps simulator-to-real deviation below the resynchronization threshold for operationally relevant timescales, yet no quantitative results are supplied on time-to-drift, deviation growth rate under non-stationary traffic, or resync latency impact. This leaves the central closed-loop assumption untested at the timescale that matters for continuous xApp operation.","section":"Evaluation section"},{"comment":"System design (RLS tuner and deviation-aware scoring): the description of the time-aware RLS correction and automatic resync trigger does not include the specific forgetting factor, gain schedule, or deviation threshold values used, nor any sensitivity analysis showing how these choices affect the duration before resync is required.","section":"System design"}],"minor_comments":[{"comment":"Abstract: the 96% accuracy figure is stated without reference to the corresponding table, figure, or experimental conditions (dataset size, traffic models, baseline comparators, or error bars).","section":"Abstract"},{"comment":"The paper would benefit from explicit comparison against a static simulator baseline or a simpler correction method to quantify the incremental benefit of the XGBoost+RLS combination.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive comments. We address each major comment below and will revise the manuscript accordingly to strengthen the evaluation and system design sections.","responses":[{"response":"We acknowledge that the evaluation does not supply explicit quantitative results on time-to-drift, deviation growth under non-stationary traffic, or resync latency. While the reported experiments achieve 96% fidelity without resynchronization during the test duration, this does not fully characterize operationally relevant timescales. In the revised manuscript we will add new experiments and analysis quantifying these metrics to substantiate closed-loop duration.","revision_made":"yes","referee_comment":"[Evaluation section] Evaluation section (energy-saving xApp experiments): the headline claims of 96% KPM accuracy and disruption-free control require that the XGBoost+RLS loop keeps simulator-to-real deviation below the resynchronization threshold for operationally relevant timescales, yet no quantitative results are supplied on time-to-drift, deviation growth rate under non-stationary traffic, or resync latency impact. This leaves the central closed-loop assumption untested at the timescale that matters for continuous xApp operation."},{"response":"We agree that the specific RLS parameters (forgetting factor, gain schedule, deviation threshold) and sensitivity analysis were not provided. In the revision we will include these values and add a sensitivity study showing their effect on resynchronization interval.","revision_made":"yes","referee_comment":"[System design] System design (RLS tuner and deviation-aware scoring): the description of the time-aware RLS correction and automatic resync trigger does not include the specific forgetting factor, gain schedule, or deviation threshold values used, nor any sensitivity analysis showing how these choices affect the duration before resync is required."}],"tokens_in":1447,"tokens_out":385,"duration_ms":24238,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that this paper builds a digital twin inside the non-RT RIC on top of ns-O-RAN-flexRIC, using XGBoost to map real KPMs to simulator configs and a time-aware RLS step to correct ongoing deviations, with a drift score that triggers resync. That two-stage loop plus the O1 streaming setup is a concrete new application for safe xApp testing.\n\nIt handles the stated problems of KPM scarcity and live-network risk in a straightforward way, and the energy-saving xApp example shows how the twin is meant to be used before pushing changes to the physical side.\n\nThe soft spot is the missing evidence. The abstract gives the 96% figure and says the xApp cuts energy without disruption, yet supplies no test setup, dataset, baseline, or error bars. The stress-test point holds: without numbers on how long the twin stays within the resync threshold under shifting traffic or configs, it is impossible to judge whether the closed loop is operationally useful. No prior-art comparison appears either.\n\nThis is for people already working on O-RAN RIC platforms or digital-twin prototypes who need a working example to build from. A reader outside that niche will get the motivation but little else.\n\nSend it to review. The framework uses real O-RAN interfaces and the pipeline is reproducible in principle, so referees can ask for the missing experiments and decide on the strength of the results.","headline":"OpenTwin packages an XGBoost-plus-RLS digital twin for O-RAN KPMs in the non-RT RIC, but the 96% accuracy and closed-loop claims rest on details the abstract does not supply.","tokens_in":2336,"tokens_out":385,"would_cite":false,"duration_ms":20600,"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":"OpenTwin creates a digital twin of O-RAN that matches real key performance metrics with up to 96 percent accuracy using XGBoost and recursive least squares correction.","keywords":["digital twin","O-RAN","KPM","xApp","energy saving","XGBoost","recursive least squares","closed loop control"],"falsifier":"Sustained KPM deviation above the resynchronization threshold for more than a brief interval after setup, observed while traffic patterns or configurations vary, would show that fidelity cannot be maintained.","tokens_in":2634,"feed_emoji":"📡","tokens_out":708,"duration_ms":23259,"temperature":0.7,"pith_summary":"The paper develops OpenTwin to overcome scarce key performance metric data from interface delays and the risks of testing AI models directly on live O-RAN networks. It deploys a simulator-based twin in the non-real-time RIC that streams measurements over the O1 interface. An XGBoost model generates simulator configuration parameters from observed time-varying behavior, while a time-aware recursive least squares tuner continuously adjusts for deviations from real measurements. A scoring system tracks fidelity and initiates resynchronization when drift occurs. Demonstrated on an energy-saving xApp, the twin validates control policies in simulation before they are applied to the physical network.","feed_headline":"Digital twin matches O-RAN KPMs with 96% accuracy","feed_subtitle":"Energy-saving xApp tests policies in simulation before live changes, avoiding disruption to operations.","key_machinery":"The two-step ML pipeline of XGBoost model for generating simulator configuration parameters from time-varying network behavior, followed by time-aware recursive least squares correction of KPM deviations, together with deviation-aware scoring that triggers resynchronization.","core_discovery":"OpenTwin is a digital twin framework built on an open-source O-RAN simulator with KPM streaming via the O1 interface. It uses a two-step approach in which an XGBoost model learns network behavior to produce simulator parameters and a time-aware recursive least squares tuner corrects deviations between twin and real-world KPMs. A deviation-aware scoring mechanism detects network drift and triggers automatic resynchronization. When paired with an energy-saving xApp, the framework tests policies safely in the virtual environment before live reconfiguration, achieving up to 96 percent accuracy in mirroring real KPMs without disrupting operations.","pith_inferences":["The same twin construction could support xApps targeting other goals such as throughput maximization or latency reduction.","If the correction steps hold across diverse traffic, the twin might operate for longer intervals before requiring manual intervention.","Extending the O1 streaming and scoring logic to additional interfaces could increase the range of metrics the twin can track."],"forward_implications":["The energy-saving xApp reduces consumption by validating policies in the twin before live application.","Control decisions reach the physical network only after virtual validation, avoiding direct disruption.","Automatic resynchronization maintains twin accuracy when drift is detected by the scoring mechanism.","KPM data becomes continuously available for xApp training without sole dependence on live interface measurements."],"fun_headline_variants":["OpenTwin achieves 96% O-RAN KPM accuracy","Digital twin tests xApps before O-RAN deployment","OpenTwin corrects KPMs with RLS in O-RAN twin","O-RAN twin detects drift and resyncs automatically"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The XGBoost configuration generator and time-aware RLS tuner together will keep simulator-to-real KPM deviation below the resynchronization threshold for operationally relevant periods even when traffic or network settings change.","fun_headline_variants_meta":{"raw":{"variants":["OpenTwin achieves 96% O-RAN KPM accuracy","Digital twin tests xApps before O-RAN deployment","OpenTwin corrects KPMs with RLS in O-RAN twin","O-RAN twin detects drift and resyncs automatically"]},"model":"grok-4.3","cost_usd":0.004741,"raw_usage":{"total_tokens":2363,"prompt_tokens":718,"num_sources_used":0,"completion_tokens":68,"cost_in_usd_ticks":47412000,"prompt_tokens_details":{"text_tokens":718,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1577,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":718,"tokens_out":68,"duration_ms":13287,"temperature":1.0,"reasoning_tokens":1577,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-30T11:51:59.284152+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Sustained KPM deviation above the resynchronization threshold for more than a brief interval after setup, observed while traffic patterns or configurations vary, would show that fidelity cannot be maintained.","supporting_citations":[],"review_version":1}