{"id":"e67e2cd6-1136-4a27-a326-f58fc2573616","arxiv_id":"2605.29939","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"CRB-guided joint optimization of sensing power and adaptive Mamba depth minimizes pose prediction error in indoor mmWave ISCC under latency and energy limits.","lead":"The paper proposes a CRB-guided resource allocation framework for indoor mmWave ISCC systems that jointly optimizes sensing power and adaptive-depth Mamba model depth to minimize human pose prediction error under constraints. A smart generalist might read it to see how sensing uncertainty, computation resources, and communication can be balanced for real-time human tracking applications.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"CRB-derived point-cloud perturbation model may not capture indoor mmWave multipath/clutter effects on actual estimation error","rationale":"Reader correctly flagged the CRB-to-perturbation link as weakest; full-text derivation (if present) does not remove the need for empirical tightness check in realistic indoor channels. Mamba adaptive-depth mapping is secondary and more readily validated by ablation. No other internal inconsistency appears from the abstract+structure.","tokens_in":1737,"tokens_out":324,"duration_ms":19985,"concrete_test":"Extract the exact CRB expression and point-cloud perturbation mapping from §III; generate synthetic indoor mmWave point clouds at 3–5 discrete sensing powers using a ray-tracing simulator that includes walls/furniture; compute empirical range RMSE and downstream pose-prediction error; compare scaling vs. the paper's CRB formula—if the empirical error deviates by >15% or changes exponent, the modeling step fails.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The framework's quantitative relationship among sensing power, model depth, and prediction error rests on characterizing range-estimation uncertainty and point-cloud perturbation via the CRB (abstract and §III presumably). CRB supplies a theoretical lower bound under simplified assumptions (e.g., single-path AWGN, known waveform), yet indoor mmWave sensing typically encounters rich multipath, clutter, beam squint, and hardware impairments that cause actual error to exceed or scale differently from the CRB. If this gap exists, the closed-form sensing-power update and the overall minimization lose their claimed grounding.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The manuscript proposes a CRB-guided resource allocation framework for indoor mmWave ISCC systems to minimize human pose prediction error under communication, latency, and energy constraints. It characterizes the impact of sensing power on range-estimation uncertainty and point-cloud perturbation via the CRB, adopts an adaptive-depth Mamba-based pose prediction model with lightweight heads after each layer, establishes a quantitative relationship among sensing power, model depth, and prediction error, formulates a joint optimization problem, and solves it with an alternating optimization algorithm yielding closed-form updates for sensing power and model depth. Simulations are reported to show reduced prediction error versus baselines.","tokens_in":1852,"tokens_out":440,"duration_ms":27604,"significance":"If the modeling assumptions hold, the work contributes a unified sensing-computation framework with closed-form AO solutions and an adaptive Mamba model, enabling efficient resource allocation for human-centric indoor applications. The explicit quantitative relationship and reproducible closed-form steps are strengths.","major_comments":[{"comment":"§III (sensing model): The point-cloud perturbation model derived from the CRB assumes single-path AWGN conditions with known waveform. Indoor mmWave environments typically involve rich multipath, clutter, beam squint, and hardware impairments, which can cause actual range-estimation error to exceed or scale differently from the CRB. This directly affects the claimed quantitative relationship among sensing power, model depth, and prediction error, as well as the grounding of the closed-form sensing-power update. The manuscript should either validate the approximation against realistic channel models or provide bounds on the deviation.","section":"§III"}],"minor_comments":[{"comment":"Abstract and simulation section: No error bars, dataset details, or specific hyperparameter values are referenced; include these for reproducibility of the reported error reductions.","section":"Abstract"},{"comment":"Notation: Ensure consistent definition of the perturbation variance term when linking CRB to the Mamba input; cross-reference the exact equation used in the optimization.","section":null}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback. We address the major comment below.","responses":[{"response":"We agree that the CRB-based point-cloud perturbation model in Section III is derived under single-path AWGN assumptions with known waveform. Indoor mmWave channels with multipath, clutter, beam squint, and hardware impairments can indeed produce range-estimation errors that exceed or scale differently from the CRB, which may affect the accuracy of the derived quantitative relationship and the closed-form sensing-power update in practical settings. The CRB is employed as a theoretical lower bound to characterize sensing uncertainty and guide resource allocation, an approach common in the ISAC literature. To address the concern, we will revise Section III to explicitly state the modeling assumptions, discuss potential deviations under realistic conditions, and derive analytical bounds on the CRB deviation for simplified multipath cases where tractable. Full validation against measured indoor channels is noted as future work. These changes will better ground the framework without altering the core contributions.","revision_made":"yes","referee_comment":"[§III] §III (sensing model): The point-cloud perturbation model derived from the CRB assumes single-path AWGN conditions with known waveform. Indoor mmWave environments typically involve rich multipath, clutter, beam squint, and hardware impairments, which can cause actual range-estimation error to exceed or scale differently from the CRB. This directly affects the claimed quantitative relationship among sensing power, model depth, and prediction error, as well as the grounding of the closed-form sensing-power update. The manuscript should either validate the approximation against realistic channel models or provide bounds on the deviation."}],"tokens_in":1352,"tokens_out":355,"duration_ms":25988,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The paper introduces a resource allocation scheme that links sensing power to point-cloud quality through the CRB, then trades that off against model depth in an adaptive Mamba predictor to cut human pose error under latency and energy limits. The alternating optimization yields closed-form updates for power and depth, which is the cleanest part of the work.\n\nWhat stands out is the explicit quantitative tie between the sensing and computation sides for this indoor mmWave setting. The adaptive-depth trick with lightweight heads after each layer is a practical way to vary compute without retraining. Simulations reportedly beat the baselines on prediction error, so the framework at least produces usable numbers.\n\nThe soft spot is the CRB step. The claim that sensing power maps cleanly to range uncertainty and point-cloud perturbation rests on the usual single-path AWGN assumptions. Indoor mmWave environments bring multipath, clutter, beam squint, and hardware effects that typically push real error well above the bound or change its scaling. The abstract does not indicate any correction for that gap, so the claimed relationship between power, depth, and error may not hold as tightly as presented.\n\nSimulation details are thin in the abstract—no error bars, dataset description, or channel model specifics—so it is hard to judge how much the gains depend on idealized conditions.\n\nThis is for people already working on integrated sensing-computation allocation for indoor tracking or edge AI. A reader in that niche can extract the solver and the modeling approach even if they adjust the sensing part themselves.\n\nIt is worth sending to referees. The problem is concrete, the method is spelled out, and the optimization is reproducible; the modeling assumptions can be stress-tested in review.","headline":"The paper gives a workable AO solver for CRB-Mamba joint allocation in indoor ISCC but the sensing model is the weak link.","tokens_in":2314,"tokens_out":411,"would_cite":false,"duration_ms":18777,"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":"CRB models the link between sensing power and model depth to minimize human pose prediction error in mmWave ISCC systems.","keywords":["ISCC","mmWave","CRB","resource allocation","human pose prediction","Mamba model","sensing power"],"falsifier":"A physical mmWave indoor testbed experiment that measures actual pose prediction errors across a range of sensing power levels and model depths and checks whether the observed errors follow the quantitative relationship derived from the CRB and adaptive model.","tokens_in":2650,"feed_emoji":"📡","tokens_out":676,"duration_ms":26242,"temperature":0.7,"pith_summary":"The paper develops a resource allocation framework for indoor millimeter-wave integrated sensing, communication, and computation systems that minimizes errors when predicting short-term human poses. It uses the Cramer-Rao bound to quantify how sensing power affects range estimation uncertainty and point-cloud quality, then ties that to an adaptive-depth Mamba model that trades computation resources for prediction accuracy. A joint optimization problem is solved via alternating optimization with closed-form updates for power and depth, subject to latency, energy, and communication limits. Sympathetic readers would care because the work shows a concrete way to allocate scarce resources for reliable real-time indoor human tracking.","feed_headline":"CRB links sensing power and depth to lower pose error","feed_subtitle":"Quantitative mapping in indoor mmWave ISCC systems enables joint optimization that cuts prediction error under latency and energy limits.","key_machinery":"CRB characterization of sensing power effects on range uncertainty and point-cloud perturbation, integrated with the adaptive-depth Mamba pose prediction model that supports inference at varying depths.","core_discovery":"The authors establish a quantitative relationship among sensing power, model depth, and prediction error by characterizing the effect of sensing power on range-estimation uncertainty and point-cloud perturbation through the CRB, adopting an adaptive-depth Mamba model with lightweight heads after each layer to map computation resources to performance, and then solving the resulting joint resource allocation problem with an alternating optimization algorithm that yields closed-form solutions for the sensing power and model depth steps.","pith_inferences":["The same CRB-to-prediction mapping could be tested on other indoor sensing tasks such as gesture or fall detection to check generality.","If the modeled relationship holds in hardware, it could support predictive scheduling of sensing and computation in time-varying indoor environments.","Extending the approach to multi-user or multi-pose scenarios would test whether the single-user optimization scales without major reformulation."],"forward_implications":["Joint allocation of sensing power and model depth under the derived relationship reduces pose prediction error while satisfying communication, latency, and energy constraints.","The alternating optimization algorithm with closed-form updates for power and depth solves the problem efficiently.","The framework outperforms baseline resource allocation methods in simulations for resource-constrained indoor human-centric applications.","The unified sensing-computation model enables explicit trade-offs between sensing resources and computation depth."],"fun_headline_variants":["CRB quantifies power and depth effects on pose error","Mamba model depth tied to sensing via CRB optimization","Resource allocation minimizes prediction error using CRB","Indoor mmWave ISCC optimizes via CRB and adaptive Mamba"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The Cramer-Rao bound accurately captures the real impact of sensing power on range-estimation uncertainty and the resulting point-cloud perturbation that drives prediction error.","fun_headline_variants_meta":{"raw":{"variants":["CRB quantifies power and depth effects on pose error","Mamba model depth tied to sensing via CRB optimization","Resource allocation minimizes prediction error using CRB","Indoor mmWave ISCC optimizes via CRB and adaptive Mamba"]},"model":"grok-4.3","cost_usd":0.005687,"raw_usage":{"total_tokens":2725,"prompt_tokens":685,"num_sources_used":0,"completion_tokens":64,"cost_in_usd_ticks":56874500,"prompt_tokens_details":{"text_tokens":685,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1976,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":685,"tokens_out":64,"duration_ms":19164,"temperature":1.0,"reasoning_tokens":1976,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-06-29T00:44:40.151365+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A physical mmWave indoor testbed experiment that measures actual pose prediction errors across a range of sensing power levels and model depths and checks whether the observed errors follow the quantitative relationship derived from the CRB and adaptive model.","supporting_citations":[],"review_version":1}