{"id":"29bb6fbc-3369-4b2a-802a-c0d831ac9675","arxiv_id":"2607.01986","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":1,"one_line_summary":"Liquid latent dynamics with disentangled degradation and condition states improve sensor forecasting RMSE to 0.2266 and degradation-state correlation to 0.5960 over GRU baselines on C-MAPSS but lag on direct RUL regression.","lead":"The paper applies liquid neural networks as latent dynamics models for turbofan engine degradation on the C-MAPSS benchmark, factorizing the latent state into degradation and operating-condition components supervised by multiple losses. A smart generalist might read it to understand whether disentangled latent models can yield more interpretable health monitoring than standard recurrent networks while still improving forecasts.","discovery_kind":"new_application","skeptic_critique":{"model":"grok-4.3","headline":"Effectiveness of the custom losses for true disentanglement without leakage remains unverified by direct checks on the learned latents.","rationale":"The reader's weakest_assumption directly identifies the same load-bearing point about the losses. The full-text description of the architecture and losses does not add an independent verification step (e.g., post-training latent correlation), so the concern stands and the UNVERDICTED status is appropriate.","tokens_in":1778,"tokens_out":287,"duration_ms":20067,"concrete_test":"After training, compute Pearson correlation and mutual information between the degradation latent vector and the condition latent vector on held-out FD002/FD004 trajectories; if either metric exceeds 0.1 on average, the decorrelation loss failed to enforce separation and the interpretability claim weakens.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that the factorization into degradation and condition components, supervised by RUL/monotonic/latent-consistency losses on one side and condition-prediction/decorrelation losses on the other, produces an interpretable degradation axis without substantial leakage. The reported RMSE drop (0.2438→0.2266) and Spearman correlation (0.5960) could arise from the liquid dynamics alone or from incomplete separation; nothing in the argument rules out the condition component still carrying degradation signal or the degradation component retaining condition dependence.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper proposes a liquid neural network as a latent dynamics model for aircraft engine health monitoring on the C-MAPSS benchmark. The latent state is factorized into degradation and condition components; the degradation component is supervised by RUL, monotonic-risk, and latent-consistency losses while the condition component uses condition-prediction and decorrelation losses. On FD001–FD004 the full model reduces sensor-forecasting RMSE from 0.2438 (GRU baseline) to 0.2266 and achieves an average state-speed Spearman correlation of 0.5960, though direct RUL regression remains stronger for the GRU.","tokens_in":1936,"tokens_out":618,"duration_ms":17945,"significance":"If the claimed separation holds, the work supplies an interpretable world model that links forecasting accuracy with an explicit degradation axis, a useful bridge between predictive-maintenance forecasting and inspectable health-state modeling. The empirical gains on the multi-condition subsets (FD002, FD004) are the most practically relevant result.","major_comments":[{"comment":"The central claim that the loss combination isolates degradation dynamics from operating-condition variation (abstract and method description) is load-bearing yet unsupported by direct verification. No ablation removes individual loss terms, no mutual-information or correlation statistics between the two latent components are reported, and no inspection of whether the condition component retains degradation signal is provided; the observed RMSE drop could therefore be produced by the liquid dynamics alone.","section":"Method / Experiments"},{"comment":"Table or figure reporting the per-subset RMSE and Spearman values (abstract) does not include standard deviations across random seeds or statistical significance tests against the GRU baseline, making it impossible to judge whether the 0.0172 RMSE reduction is reliable or within noise.","section":"Experiments"},{"comment":"The paper states that the degradation component forms a clearer temporal axis (Spearman 0.5960) but supplies no quantitative comparison of this metric for the GRU baseline or for ablated versions of the proposed model, so the contribution of the factorization versus the liquid transition itself cannot be isolated.","section":"Experiments"}],"minor_comments":[{"comment":"The abstract notes that direct RUL regression is stronger for the GRU but gives no numerical values; adding those numbers would allow readers to weigh the trade-off between interpretability and regression accuracy.","section":"Abstract"},{"comment":"Notation for the liquid transition function and the two latent subspaces should be introduced once with consistent symbols rather than re-defined in the loss sections.","section":"Method"}],"recommendation":"major_revision","confidential_remarks":"The manuscript reports concrete numbers on a public benchmark but the training protocol, hyper-parameter search, and statistical details are not described in the provided text; this limits verification of the numerical claims. The work is within scope for an ML-for-prognostics venue."},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback. We address each major comment below and will revise the manuscript accordingly to provide stronger empirical support for the disentanglement claims and statistical robustness of the results.","responses":[{"response":"We agree that the manuscript does not provide direct ablations or quantitative verification (such as mutual information or correlations) of the separation between the degradation and condition latent components. While the loss functions are explicitly designed to target this separation, the absence of these checks means the contribution of the factorization cannot be fully isolated from the liquid dynamics. In the revised manuscript we will add an ablation study removing each loss term, report mutual information and Pearson/Spearman correlations between the two latent components, and inspect the condition component for residual degradation signal via correlation with RUL.","revision_made":"yes","referee_comment":"[Method / Experiments] The central claim that the loss combination isolates degradation dynamics from operating-condition variation (abstract and method description) is load-bearing yet unsupported by direct verification. No ablation removes individual loss terms, no mutual-information or correlation statistics between the two latent components are reported, and no inspection of whether the condition component retains degradation signal is provided; the observed RMSE drop could therefore be produced by the liquid dynamics alone."},{"response":"We acknowledge that variability across seeds and statistical significance testing are necessary to assess whether the reported RMSE improvement is reliable. The revised manuscript will include results aggregated over multiple random seeds, with standard deviations reported in the tables, and paired statistical tests (e.g., t-tests) against the GRU baseline.","revision_made":"yes","referee_comment":"[Experiments] Table or figure reporting the per-subset RMSE and Spearman values (abstract) does not include standard deviations across random seeds or statistical significance tests against the GRU baseline, making it impossible to judge whether the 0.0172 RMSE reduction is reliable or within noise."},{"response":"The current manuscript reports the Spearman correlation only for the full proposed model. To allow isolation of the factorization's contribution, the revision will compute and report the same Spearman metric on the GRU baseline (using its hidden state) as well as on ablated versions of the proposed model.","revision_made":"yes","referee_comment":"[Experiments] The paper states that the degradation component forms a clearer temporal axis (Spearman 0.5960) but supplies no quantitative comparison of this metric for the GRU baseline or for ablated versions of the proposed model, so the contribution of the factorization versus the liquid transition itself cannot be isolated."}],"tokens_in":1475,"tokens_out":554,"duration_ms":24978,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that this paper applies liquid neural networks to latent dynamics for turbofan degradation on C-MAPSS, factorizing the state into degradation and condition components with custom losses, and gets a modest improvement in sensor forecasting RMSE from 0.2438 to 0.2266 while achieving a 0.596 Spearman correlation for the degradation state.\n\nWhat is new is the combination of liquid transition models with explicit latent factorization and the suite of disentanglement losses tailored to separate health evolution from operating conditions. The liquid NN part builds on existing work, but the application to this safety-relevant aviation task with the added losses is the fresh element. It does well by providing concrete results on a public benchmark against a GRU baseline, with larger gains on the multi-condition subsets, and by highlighting that the representation works better as an interpretable world model than for direct RUL regression.\n\nThe soft spots are that the disentanglement effectiveness isn't directly verified. The losses aim to prevent leakage, but without ablations or checks on the learned latents, it's possible the gains come from the dynamics alone or that separation is incomplete. The RUL prediction being weaker than the baseline is a clear limitation for the core prognostics use case. Training protocols and hyperparameter details would need checking in the full paper to confirm the numbers hold up.\n\nThis is for people in ML for prognostics, particularly those interested in interpretable latent models for time-series health monitoring. A reader focused on aviation applications or disentangled representations would get value from the benchmark results.\n\nI would send this to peer review because it has a clear empirical contribution on a standard dataset and the claims are not overstated.","headline":"Liquid NNs with factored latents and custom losses give small sensor-forecasting gains on C-MAPSS but the separation claim lacks direct checks and RUL prediction lags the GRU baseline.","tokens_in":2408,"tokens_out":422,"would_cite":false,"duration_ms":32460,"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":"Liquid neural networks evolve a factorized latent state to forecast turbofan sensors more accurately while exposing a clearer degradation axis.","keywords":["liquid neural networks","latent dynamics","turbofan degradation","prognostics","disentangled representations","C-MAPSS","sensor forecasting","interpretable models"],"falsifier":"An ablation removing the condition decorrelation loss that leaves forecasting RMSE unchanged or higher while dropping the degradation-state Spearman correlation below 0.5960 would show the separation is not effective.","tokens_in":2679,"feed_emoji":"✈️","tokens_out":714,"duration_ms":22923,"temperature":0.7,"pith_summary":"The paper tests liquid neural networks as latent dynamics models for aircraft engine health on the C-MAPSS benchmark. A history window is encoded into a latent state that a liquid transition evolves, then decoded into future sensor values. The state is split into degradation and condition parts; the degradation part is trained with remaining useful life, monotonic risk, and consistency losses while the condition part uses prediction and decorrelation losses. This yields lower sensor forecasting error than a GRU baseline, especially on multi-condition data, and produces a degradation state whose value tracks degradation speed more closely. The resulting representation works better as an inspectable model of health dynamics than as a direct lifetime predictor.","feed_headline":"Liquid states disentangle engine degradation from conditions","feed_subtitle":"The model lowers sensor forecast RMSE to 0.2266 and raises degradation-state correlation to 0.596 on C-MAPSS data.","key_machinery":"Liquid transition model applied to a latent state factorized into degradation and condition components and trained with combined forecasting, RUL, monotonic risk, latent-consistency, condition prediction, and decorrelation losses.","core_discovery":"Encoding sensor history into a latent state, evolving that state with a liquid transition model, and decoding future observations produces improved sensor forecasting RMSE of 0.2266 versus 0.2438 for a GRU baseline across FD001-FD004. Factorizing the latent state into a degradation component supervised by RUL, monotonic risk, and latent-consistency losses plus a condition component supervised by prediction and decorrelation losses creates a degradation state with average Spearman correlation 0.5960 to degradation speed. Direct remaining useful life regression remains stronger for the GRU baseline, so the liquid model functions more effectively as an interpretable world model for degradation","pith_inferences":["The same factorization approach could be applied to other multivariate time-series tasks where hidden dynamics mix with external covariates.","If the degradation state remains stable under distribution shift, it could support anomaly detection by tracking deviations from expected trajectories.","The method points toward hybrid models that retain forecasting strength while adding explicit state inspection for maintenance decisions."],"forward_implications":["Sensor forecasting accuracy gains are largest on the multi-condition subsets FD002 and FD004.","The degradation component of the latent state forms a clearer temporal axis aligned with health decline.","The model serves as an interpretable world model for degradation dynamics rather than a calibrated lifetime regressor.","Liquid latent dynamics can connect predictive maintenance forecasting with inspectable health-state tracking."],"fun_headline_variants":["Liquid states factor turbofan degradation from operating conditions","Liquid model reaches 0.2266 RMSE for turbofan sensor forecasts","Factorized liquid states reach 0.596 degradation correlation","Liquid dynamics separate engine degradation from conditions"],"cache_read_input_tokens":64,"weakest_assumption_plain":"The custom losses separate health evolution from operating-condition variation without substantial information leakage or loss of forecasting performance.","fun_headline_variants_meta":{"raw":{"variants":["Liquid states factor turbofan degradation from operating conditions","Liquid model reaches 0.2266 RMSE for turbofan sensor forecasts","Factorized liquid states reach 0.596 degradation correlation","Liquid dynamics separate engine degradation from conditions"]},"model":"grok-4.3","cost_usd":0.00663,"raw_usage":{"total_tokens":3128,"prompt_tokens":737,"num_sources_used":0,"completion_tokens":62,"cost_in_usd_ticks":66299500,"prompt_tokens_details":{"text_tokens":737,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2329,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":737,"tokens_out":62,"duration_ms":15701,"temperature":1.0,"reasoning_tokens":2329,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-07-03T17:15:39.852921+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"An ablation removing the condition decorrelation loss that leaves forecasting RMSE unchanged or higher while dropping the degradation-state Spearman correlation below 0.5960 would show the separation is not effective.","supporting_citations":[],"review_version":1}