{"id":"b83406e2-ee61-4b2b-856f-b6246239f7d5","arxiv_id":"2602.18473","paper_version":2,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"CoTAR replaces attention with a centralized core-token MLP that aggregates and redistributes channel information to match the centralized structure of medical time series signals, yielding accuracy gains and linear complexity.","lead":"The paper proposes CoTAR, a module using one global core token to aggregate and redistribute information across channels in medical time series instead of standard decentralized attention. This targets better modeling of synchronized signals like EEG and ECG while cutting compute costs.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"Single core token may bottleneck nuanced inter-channel dependencies in MedTS","rationale":"The reader's weakest assumption directly identifies the same point: whether the core-token proxy preserves necessary global patterns without loss. Because the review was abstract-only, the concrete_test above supplies the missing verification step that would either confirm or refute the claim on the exact dataset where gains are reported. No other internal inconsistency is visible from the given description.","tokens_in":1784,"tokens_out":284,"duration_ms":17754,"concrete_test":"On the APAVA benchmark, train a CoTAR variant that uses k=4 independent core tokens (each aggregating a random partition of channels) instead of k=1; if test accuracy falls by >3% relative to the reported single-core result while keeping total parameter count matched, the single-proxy assumption is the limiting factor.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that routing all interactions through one global core token (via aggregation then redistribution) fully preserves the synchronization patterns that decentralized attention would capture directly. This holds only if the core token's representation is lossless for the relevant global statistics; a simple learned projection or pooling step into a fixed-size vector can discard channel-specific phase or amplitude variations that matter for diagnosis. The abstract provides no formal argument or capacity analysis showing why a single proxy suffices where pairwise attention fails.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper claims that medical time series (MedTS) signals such as EEG and ECG exhibit centralized patterns that mismatch the decentralized nature of standard Transformer attention, leading to poor capture of global synchronization. It introduces CoTAR, an MLP-based module that replaces attention with a global core token acting as a proxy for aggregation and redistribution of inter-token information. This design is argued to better align with MedTS properties while reducing complexity from quadratic to linear. Experiments on five benchmarks are reported to show superiority, with up to 11.6% improvement on APAVA, 33% memory usage, and 20% inference time relative to prior state-of-the-art.","tokens_in":1857,"tokens_out":454,"duration_ms":38166,"significance":"If validated, the work offers a concrete architectural alternative for MedTS that prioritizes centralized aggregation, with potential efficiency gains that could benefit real-time clinical applications. The public release of code and training scripts strengthens reproducibility and allows direct testing of the linear-complexity claim.","major_comments":[{"comment":"Abstract: the central claim that routing all interactions through a single global core token 'fully preserves' synchronization patterns without loss is load-bearing but unsupported; no capacity analysis, formal argument, or ablation demonstrates why a fixed-size proxy suffices where pairwise attention fails, especially given the risk of discarding channel-specific phase/amplitude variations.","section":"Abstract"},{"comment":"Abstract: the reported gains (e.g., 11.6% on APAVA) and efficiency numbers rest on experimental assertions with no implementation details, ablation studies, or error analysis supplied, so the superiority claim cannot be assessed for robustness or confounding factors.","section":"Abstract"}],"minor_comments":[{"comment":"The abstract states 'Code and all training scripts are available' but provides no commit hash or exact reproduction instructions, which would aid verification.","section":"Abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript appears to rest heavily on unverified experimental claims; if the full text contains only the same level of detail as the abstract, the work may require substantial additional experiments before acceptance."},"author_rebuttal":{"model":"grok-4.3","summary":"We appreciate the referee's thorough review and constructive feedback on our manuscript. We have carefully considered each major comment and provide point-by-point responses below. We believe the revisions will address the concerns raised and strengthen the paper.","responses":[{"response":"We thank the referee for highlighting this important point. While our experiments on multiple MedTS benchmarks empirically demonstrate that the core token effectively captures global synchronization without significant loss (as evidenced by performance gains), we acknowledge the absence of a formal capacity analysis in the current manuscript. In the revised version, we will include a theoretical discussion on the information preservation properties of the core token and additional ablations varying the core token size to address concerns about discarding channel-specific variations.","revision_made":"yes","referee_comment":"[Abstract] Abstract: the central claim that routing all interactions through a single global core token 'fully preserves' synchronization patterns without loss is load-bearing but unsupported; no capacity analysis, formal argument, or ablation demonstrates why a fixed-size proxy suffices where pairwise attention fails, especially given the risk of discarding channel-specific phase/amplitude variations."},{"response":"We agree that the abstract is limited in space and does not include implementation details or ablations. The full manuscript provides details on the experimental setup, and the code is publicly released for reproducibility. To strengthen the claims, we will add ablation studies on key components and report standard deviations across multiple runs in the revised manuscript to allow assessment of robustness.","revision_made":"yes","referee_comment":"[Abstract] Abstract: the reported gains (e.g., 11.6% on APAVA) and efficiency numbers rest on experimental assertions with no implementation details, ablation studies, or error analysis supplied, so the superiority claim cannot be assessed for robustness or confounding factors."}],"tokens_in":1420,"tokens_out":390,"duration_ms":39387,"standing_objections":[]},"desk_editor":{"model":"grok-4.3","letter":"The main takeaway is that this paper replaces standard decentralized attention with CoTAR, a centralized MLP module that routes all inter-token interactions through one global core token for aggregation then redistribution. It targets the mismatch between typical transformer attention and the synchronized, centralized patterns in signals like EEG and ECG, while dropping complexity from quadratic to linear. They report up to 11.6% gains on APAVA and big resource savings (33% memory, 20% inference time) across five benchmarks, with code released.","headline":"CoTAR swaps attention for a single core token to get linear complexity on MedTS, with reported efficiency wins, but the case that one proxy loses nothing important is thin.","tokens_in":2321,"tokens_out":180,"would_cite":false,"duration_ms":33914,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[{"relation":"echoes","rs_module":"IndisputableMonolith/Foundation/AbsoluteFloorClosure.lean","rs_theorem":"absolute_floor_iff_bare_distinguishability","paper_passage":"CoTAR introduces a global core token that serves as a proxy to facilitate inter-token interactions, thereby enforcing a centralized aggregation and redistribution strategy"},{"relation":"unclear","rs_module":"IndisputableMonolith/Foundation/AlexanderDuality.lean","rs_theorem":"alexander_duality_circle_linking","paper_passage":"MedTS signals are inherently centralized, whereas the Transformer’s attention is decentralized"}],"headline":"CoTAR centralized proxy token loosely echoes RS central forcing but is orthogonal to the framework","alignment":"orthogonal","rationale":"The paper's central machinery (global core token for aggregation-redistribution replacing decentralized attention) draws an analogy to centralized biological sources (brain/heart) and star-shaped systems, which superficially parallels RS's derivation of centralized structures from a single distinction (reality_from_one_distinction, AbsoluteFloorClosure). However, it introduces no J-cost, φ-ladder, 8-tick periodicity, or parameter-free constant derivations; the design is a practical MLP engineering choice for linear complexity in MedTS transformers. No overlap with RS theorems on cost functions, arithmetic recovery, or spacetime emergence. Domain (cs.LG medical time-series) lies outside RS scope.","tokens_in":62735,"confidence":"low","tokens_out":331,"duration_ms":16660,"cache_read_input_tokens":38528,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"A global core token can replace decentralized attention to better match the centralized structure of medical time series signals.","keywords":["medical time series","transformer attention","centralized aggregation","core token","EEG","ECG","channel dependencies","computational efficiency"],"falsifier":"A controlled experiment where a standard transformer using full decentralized attention achieves equal or higher accuracy than CoTAR on the APAVA or similar MedTS benchmarks, with no increase in memory usage, would refute the central claim.","tokens_in":2696,"feed_emoji":"🧠","tokens_out":731,"duration_ms":46620,"temperature":0.7,"pith_summary":"Medical time series data such as EEG and ECG show strong centralized patterns across channels, including global synchronization and unified waveforms. Standard transformer attention lets every token interact directly with every other token in a decentralized way, which creates a structural mismatch with this data property. The paper introduces CoTAR, an MLP-based module that routes all interactions through one global core token acting as a proxy: it aggregates information from all tokens then redistributes the result back out. This enforces centralized flow, captures channel dependencies more effectively, and reduces complexity from quadratic to linear scaling. Experiments on five benchmarks show accuracy gains of up to 11.6 percent alongside large cuts in memory and inference time.","feed_headline":"Core token replaces attention for medical time series","feed_subtitle":"CoTAR routes interactions through one global proxy to match centralized MedTS patterns, using 33% memory and 20% time of prior models.","key_machinery":"The global core token in CoTAR, which aggregates information from all tokens and redistributes a unified representation to enforce centralized inter-token interactions.","core_discovery":"The paper claims that the decentralized design of standard attention mechanisms fundamentally mismatches the centralized nature of medical time series signals, which require global synchronization and unified patterns across channels. CoTAR solves this by introducing a single global core token that aggregates information from all input tokens and then redistributes a unified representation back to them, replacing direct token-to-token interactions with a centralized aggregation-redistribution strategy. This alignment with the data structure also reduces computational complexity from quadratic to linear while preserving the ability to model temporal dependencies within channels.","pith_inferences":["The same core-token design may apply to other centralized multivariate time series outside medicine, such as multi-sensor industrial monitoring.","Hybrid models could combine the centralized core with selective decentralized attention for data that mixes global and local patterns.","The linear scaling opens the door to real-time processing of very long recordings that quadratic attention cannot handle.","The approach suggests examining whether other data domains with centralized structure would benefit from similar proxy-token mechanisms."],"forward_implications":["Channel dependencies in EEG and ECG data are modeled more effectively through centralized aggregation than through pairwise attention.","Computational complexity drops from quadratic to linear in sequence length, enabling longer time series inputs.","Memory consumption falls to roughly one-third and inference time to one-fifth of prior state-of-the-art transformer models.","Accuracy improves by up to 11.6 percent on the APAVA dataset while maintaining temporal modeling within channels.","The centralized module can directly replace attention layers in existing transformer architectures for MedTS tasks."],"fun_headline_variants":["Decentralized Attention Fails Centralized MedTS Signals","Core Token Centralizes Transformer Interactions for MedTS","CoTAR Routes MedTS via Centralized Core Token","Rethinking Transformers With Centralized Core Token for MedTS"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"Medical time series signals have an inherently centralized structure such that routing all channel interactions through one core token proxy captures the necessary global synchronization patterns without loss.","fun_headline_variants_meta":{"raw":{"variants":["Decentralized Attention Fails Centralized MedTS Signals","Core Token Centralizes Transformer Interactions for MedTS","CoTAR Routes MedTS via Centralized Core Token","Rethinking Transformers With Centralized Core Token for MedTS"]},"model":"grok-4.3","cost_usd":0.006195,"raw_usage":{"total_tokens":2961,"prompt_tokens":751,"num_sources_used":0,"completion_tokens":60,"cost_in_usd_ticks":61949500,"prompt_tokens_details":{"text_tokens":751,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":2150,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":751,"tokens_out":60,"duration_ms":15596,"temperature":1.0,"reasoning_tokens":2150,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-16T06:07:52.647611+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"A controlled experiment where a standard transformer using full decentralized attention achieves equal or higher accuracy than CoTAR on the APAVA or similar MedTS benchmarks, with no increase in memory usage, would refute the central claim.","supporting_citations":[],"review_version":1}