Pith. sign in

REVIEW 3 major objections 4 minor 54 references

This paper claims that the clinical act for arrhythmias lies between naming a rhythm and deciding what to do, and presents a multi-agent system that spans that span, from raw ECG and pulse windows to a patient-level rhythm profile and guide

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 02:42 UTC pith:Q6TZT7KT

load-bearing objection Solid multi-agent pipeline for a genuinely new task, but the 'wins on every axis' claim is partly an artifact of an evaluation rubric that rewards the guideline-citation machinery the system itself contains. the 3 major comments →

arxiv 2607.25340 v1 pith:Q6TZT7KT submitted 2026-07-28 cs.AI cs.CL

Cardiologent: Multi-Agent Clinical Decision Support for Patient-Level Arrhythmia Assessment, Urgency, and Management

Sukju Oh , Moo-Yong Rhee , Jae-Sik Jang , Sukkyu Sun This is my paper
classification cs.AI cs.CL
keywords patient-level arrhythmia decision supportmulti-agent systemECG and PPG monitoringrhythm profile assemblyguideline-concordant managementauditable clinical decisionsLLM-as-judge evaluationwearable arrhythmia surveillance
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tries to establish that patient-level arrhythmia decision support—deciding what a monitored record means for this patient and what to do about it—is a distinct task that can be built end to end, and that its system, Cardiologent, is the first to span it from raw signals to auditable management decisions. The driving observation is that identical arrhythmia signals can warrant opposite decisions depending on the patient, so the system must assemble window-level readings into a patient rhythm profile and read that profile against the patient's own data and clinical guidelines. The authors claim Cardiologent scores highest on every evaluated axis—integrated diagnosis, clinical significance, urgency, and management—under both blinded cardiologists and a large-scale LLM judge, and that its window-level classification macro-F1 (0.449) is roughly double the strongest baseline's. A sympathetic reader would care because earlier systems stop either at naming the rhythm or at interpreting device-supplied findings, leaving out the judgement a clinician would act on.

Core claim

At the paper's center is the claim that the clinical act is the span between naming a rhythm and knowing what to do, and that this span can be mechanized. Cardiologent is a two-level multi-agent system: at the window level, separate agents read a single ECG lead and a wearable PPG through deterministic measurement tools—heart rate, R–R irregularity, QRS width, P-wave presence, pulse morphology—so every claim is tied to a measured quantity; the two readings are reconciled by a debate in which analysts, a devil's advocate, and a referee grant authority feature-by-feature to the signal that can see the feature. At the patient level, the window verdicts are aggregated into a rhythm profile and o

What carries the argument

The central object is Cardiologent's two-level pipeline. The load-bearing identity is the per-feature authority assignment during fusion: rather than giving one sensor fixed precedence, a referee decides which sensor's measured evidence is trustworthy for each feature—the ECG for QRS width and P waves, the PPG for pulse regularity and perfusion—so a correct minority reading can survive. The other load-bearing mechanism is the patient-level critic, which holds every diagnosis, significance claim, and management recommendation to a retrieved and cited guideline passage, deleting or rewriting claims that fail that test. The same analyst–devil's advocate–referee debate primitive appears at both

Load-bearing premise

The load-bearing premise is that the scoring instrument—an LLM-judge rubric whose grade descriptions are written from the same guidelines the system retrieves, with human validation on only 20 patients and two cardiologists—captures real clinical decision quality. If the scores mostly measure closeness to those specific guidelines, the 'highest on every axis' claim does not transfer to clinical use.

What would settle it

A blinded, multi-centre evaluation in which cardiologists score management decisions on 100+ patients without knowing the system's identity; if Cardiologent is not ranked first, or if swapping the guideline corpus for a different authoritative set changes the ranking, the central claim is refuted.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If correct, continuous monitoring no longer needs a separate device that pre-detects arrhythmias: the same system can read raw ECG and pulse windows, assemble the patient's rhythm profile, and return a management recommendation.
  • A wearable could run the pulse-only agent alone for low-cost surveillance and escalate to a combined ECG-plus-pulse read when the pulse reading warrants it, without changing systems.
  • The reported advantage concentrates on rare, high-risk rhythms—ventricular tachycardia, AV block, sinus node dysfunction—where general models read near zero; this is where a deployed system would add the most.
  • Because every conclusion is tied to a retrieved, cited guideline passage, a clinician can audit and overturn the system's decision rather than accept an unexplained alert.
  • The patient-level results imply that evaluating report fluency would miss the real failure mode: general models are fluent but confidently wrong about the patient, so decision-level scoring is necessary.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Inference: If the rubric is indeed measuring decision quality, the two-level design—grounded per-signal reading, feature-level arbitration, citation-checked reasoning—should transfer to other continuously monitored signals and other guideline-driven care decisions; that transfer is not tested in the paper.
  • Inference: The ranking may be partly an artifact of writing rubric anchors from the same guidelines the system retrieves; swapping in a different authoritative guideline set and re-ranking would show whether the advantage is about following guidelines generally or about matching this particular corpus.
  • Inference: The evaluation is retrospective and from one institution's monitoring data, so the implicit next claim—that the system improves live clinical decisions—remains untested; a prospective deployment study would be the natural sequel.
  • Inference: The debate mechanism suggests a more general principle: when two sensors have complementary physical access to a feature, authority should be assigned per feature rather than by a fixed sensor hierarchy.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper defines patient-level arrhythmia decision support as a task spanning window-level classification (single-lead ECG and PPG) through patient-level integrated diagnosis, clinical significance, and urgency/management. Cardiologent comprises per-signal LLaVA-7B specialists augmented with deterministic measurement tools and debate-based fusion; at patient level it aggregates window verdicts into a rhythm profile, retrieves guideline passages with PMIDs, reasons over the profile and metadata, and applies a critic that checks claims against retrieved citations. The authors evaluate window classification (macro-F1), reasoning probes (T1–T3), grounding (IoU/MCC), and patient-level P1–P3 scoring using DeepSeek-V4-Pro as an LLM judge on 65 patients and two blinded cardiologists on a 20-patient subset. They report that Cardiologent ranks first on every axis under both raters, that judge–cardiologist ICC is comparable to human–human agreement, and that window macro-F1 is 0.449 versus 0.255 for the strongest baseline.

Significance. If the evaluation is accepted, the contribution is significant: the system is, to my knowledge, the first to assemble a patient-level arrhythmia assessment from raw ECG+PPG through to a guideline-cited management decision, and it does so with a carefully motivated tool-and-debate architecture. The paper has real strengths: a patient-disjoint split, a 3,000-window evaluation sample, bootstrap confidence intervals, per-axis anchored rubrics, a hard-label diagnosis check that cannot be inflated by the rubric, blinded cardiologist ratings, and a promise of released code/data. The window-level results are credible and the ablation of the agent layer over the specialist is informative. However, the patient-level evaluation, which carries the headline 'scores highest on every axis,' has load-bearing weaknesses in the construction of the P3 rubric, the symmetry of information between system and baselines, and the size of the human validation.

major comments (3)
  1. [Scoring Rubrics (P3), Patient-Level Pipeline, Baseline Prompts] The P3 management rubric awards grade 4 to a plan that is 'specific to the rhythm and patient, guideline-concordant, with a correct citation' and grade 3 penalizes a 'loose' citation. Cardiologent alone is equipped with a curated guideline corpus whose passages carry PMIDs (Patient-Level Pipeline, Retrieval) and a critic that enforces that management claims cite a retrieved PMID (Critic bullet 'Citations'). Baselines receive no guideline text or retrieval tool; their prompt asks only for a JSON P3 grounded in waveform and metadata. Table 4 P3 therefore conflates decision quality with access to the citation mechanism. Since the paper's headline is 'scores highest on every axis,' this is load-bearing. The authors should either give baselines the same retrieval resource, rate under a rubric that does not reward citations, or ablate retrieval/critic to show the margin is not explained by cit
  2. [Table 13, Figure 4, Limitations] The at-scale judge is validated on two cardiologists and twenty patients. The claim that judge–cardiologist agreement (ICC 0.74, 0.66) 'matches' human–human agreement (0.67) is based on 20 patients; per-task confidence intervals are wide, e.g., P3 human–human ICC = 0.480 ± 0.170, Judge↔R1 = 0.518 ± 0.172, Judge↔R2 = 0.488 ± 0.166. This is compatible with the judge being a poor proxy, and the 20-patient subset is too small to rule out that agreement is driven by surface features such as citations. The paper itself states in Limitations: 'The evaluation is also retrospective—we score the decision itself, not whether acting on it would have changed the patient's course.' As the sole clinical validation for a strong comparative claim, this needs substantially more data or appropriately hedged wording.
  3. [Scoring Rubrics, 'Where the anchors come from'] The patient-level anchors are explicitly 'written against clinical practice' using the 2023 AF/CHA2DS2-VASc, 2017 VT/SCD, and 2018 bradycardia guidelines — the same guidelines that constitute the curated retrieval corpus and the same citations the critic enforces. The evaluation instrument is therefore aligned with the system's design: P2 context rewards patient-feature-to-risk reasoning that retrieval is built to supply, and P3 management rewards the guideline-cited plan that retrieval and critic are built to produce. This is not a derivation circularity, but it means the comparison measures closeness to the specific guidelines Cardiologent was engineered to cite. At minimum, the paper should show that baselines fail on these axes even when provided the same guidelines, or that the judge's ordering persists for clinical correctness independent of citation presence.
minor comments (4)
  1. [Experiments, baseline list] Qwen2.5-VL-72B is cited as '(?)' in the baseline enumeration; this is a missing reference placeholder and should be completed.
  2. [Method, ECG and PPG Agents] Typo: 'Around its it deterministic tools' should read 'Around it are deterministic tools.'
  3. [Table 9 caption] The caption says 'Setting both is nine-class (ECG specialist)', but the 'Cardiologent' column in that table is the full fused window system, not the ECG specialist alone; please clarify the column semantics to avoid confusion.
  4. [Figure 3 caption] The caption says 'seven on the ECG, nine on the PPG' for tools; the main text says the ECG agent has seven tools and the PPG agent has nine. This is correct but easy to misread as class counts; consider rewording.

Circularity Check

1 steps flagged

P3 rubric's citation anchor is satisfiable by Cardiologent's RAG/critic by construction, while baselines receive no guideline corpus; central window-level and hard-label results are independent, so circularity is partial

specific steps
  1. other [Scoring Rubrics (P3 management); Method > Patient Reasoning > Retrieval/Critic; Experiments > Baseline Prompts]
    ""4 specific to the rhythm and patient, guideline-concordant, with a correct citation." "Each retrieved passage carries its citation, so the evidence it supplies arrives already referenced"; critic: "Citations—every risk or management claim must cite a PMID that appears in the retrieved set". Baselines are "given the same information the task provides... What our system adds is not more information but its orchestration over that shared input.""

    The top P3 grade requires "a correct citation." Cardiologent's RAG supplies PMIDs with every retrieved passage and its critic rewrites management claims to cite a retrieved PMID. Baselines are prompted only to ground statements in waveforms/metadata and receive no guideline corpus or retrieval tool, so they cannot satisfy the citation condition by construction. The P3 management margin therefore measures access to the guideline corpus as much as clinical decision quality, making the "scores highest on every axis" claim partially built into the scoring instrument rather than derived from the system's decisions.

full rationale

This is an empirical system paper, not a derivation, so classic prediction-equals-fit circularity does not arise. The window-level classification (macro-F1 0.449 vs 0.255), grounding MCC, and patient-level hard-label diagnosis accuracy (0.48 exact vs 0.17) are measured against external ground truth on a patient-disjoint split of VitalDB and are independent of the system's design. The LLM-judge agreement with two blinded cardiologists (ICC 0.74, 0.66 vs human-human 0.67) is a real external anchor, though the 20-patient human sample is small. The main circular element is confined to the patient-level P3 management axis: the top rubric grade requires "a correct citation"; Cardiologent's RAG retrieves guideline passages with embedded PMIDs and its critic enforces that every management claim cite a retrieved PMID, while baselines are explicitly given no guideline corpus or retrieval tool and are prompted only to ground statements in waveforms and metadata. The citation condition is thus satisfiable by Cardiologent by construction and effectively unavailable to baselines, so the P3 comparison — and the headline "scores highest on every axis" insofar as it includes P3 — is partly an artifact of the evaluation instrument rather than of clinical decision quality. The paper's own Limitations concede the evaluation is retrospective and does not measure patient outcomes. This is a partial evaluation circularity, not a derivation collapse; the core detection and diagnosis claims stand on external benchmarks.

Axiom & Free-Parameter Ledger

5 free parameters · 8 axioms · 0 invented entities

The central results depend heavily on trained weights, hand-chosen preprocessing, and hand-authored evaluation components, not on a parameter-free derivation. The main fitted/free elements are the two-stage specialist models, the class-dependent window stride, the curated guideline corpus, and the critic rule set. No new physical entities are introduced.

free parameters (5)
  • Stage-1 TimeMAE encoder weights (ECG/PPG specialists) = not enumerated (network weights)
    Trained by masked reconstruction on VitalDB windows; the representation used for classification and reasoning is fitted to the training distribution.
  • Stage-2 LLaVA-7B specialist weights = not enumerated
    Fine-tuned on QA targets generated from ground-truth labels; every window reading depends on these fitted weights.
  • Class-dependent sliding stride = 1-peak stride for rare classes (~92% overlap), 3-peak for common (~75%)
    Chosen by hand to counteract class imbalance; changes the effective training distribution and the sampled evaluation windows.
  • Curated guideline retrieval base = about twenty guideline snippets
    Manually selected local corpus used in evaluation; the set bounds which citations the system can retrieve, so it influences P2/P3 scores.
  • Critic rule set = five hand-specified rules (grounding, citations, anticoagulation, redundant tests, class specificity)
    Manually authored post-processing rules; directly edits management outputs before scoring.
axioms (8)
  • domain assumption VitalDB Arrhythmia Database labels are accurate enough to serve as ground truth for training and evaluation.
    All window and patient labels derive from anesthesiologist annotations (κ=0.930); any label noise propagates to specialist training and every reported score.
  • domain assumption ECG-derived rhythm labels can be transplanted to the PPG by pulse-arrival-time alignment.
    Dataset section; the PPG-only seven-class task and the PPG agent's training depend on this transfer.
  • domain assumption A 10-second, 12-beat window is sufficient to assign the nine rhythm classes.
    Window definition in Dataset Details; rare rhythms like VT may need longer context.
  • domain assumption The LLM judge and rubric are a valid measure of clinical decision quality.
    Scoring Rubrics; supported only by ICC on 20 patients, not by outcome data.
  • domain assumption Lead II ECG and PPG from intraoperative VitalDB represent the continuous-monitoring/wearable deployment.
    Introduction/Limitations; the paper acknowledges this is unverified for wearables.
  • domain assumption Guideline passages retrieved are the correct standard of care for these patients.
    Patient-Level Pipeline; management scores assume 2023 AF/CHA2DS2-VASc, 2017 VT, 2018 bradycardia guidelines apply to the intraoperative setting.
  • domain assumption Multi-agent debate improves reasoning correctness.
    Window Fusion cites Du et al. 2023, Liang et al. 2024; not formally proven.
  • domain assumption Ground-truth-anchored QA generation produces faithful training targets and golds.
    Specialist QA / Reasoning benchmark; labels are used as anchors to generate the text the specialist is trained to produce.

pith-pipeline@v1.3.0-alltime-deepseek · 28751 in / 15984 out tokens · 145943 ms · 2026-08-01T02:42:06.976765+00:00 · methodology

0 comments
read the original abstract

The same episode of atrial fibrillation is a minor finding in a healthy adult and grounds for anticoagulation in an elderly patient with hypertension: identical signal, opposite decision. Naming the rhythm is only the start; what determines a patient's outcome is the judgement that follows -- what the arrhythmia is across the whole record, what it means for this patient, and what should be done about it. Recent work pairing large language models with the ECG stops short of this, reading one recording without assembling a patient-level finding; and agentic systems built around it either receive the arrhythmia a device has already detected or target a different diagnostic task, stopping before the decision this task requires. We formulate patient-level arrhythmia decision support as a task and present Cardiologent, a multi-agent system that spans it from detection to decision. An agent for each signal -- a single ECG lead and the photoplethysmogram a wearable acquires -- grounds its window reading in measured features rather than a bare label; the readings are assembled into the patient's rhythm profile and, with the patient's own data, reasoned against clinical guidelines retrieved for the case, with a critic checking each conclusion against the guideline it cites. We evaluate the clinical decision rather than the report, across integrated diagnosis, clinical significance, and urgency and management. Cardiologent scores highest on every axis, first on every patient-level task under both cardiologists and an at-scale LLM judge -- whose agreement with the cardiologists (ICC 0.74, 0.66) matches theirs with each other (0.67). Because each conclusion traces to a cited guideline and is validated against expert cardiologists, it yields decisions a clinician can audit rather than act on blindly -- a step toward use in continuous monitoring.

Figures

Figures reproduced from arXiv: 2607.25340 by Jae-Sik Jang, Moo-Yong Rhee, Sukju Oh, Sukkyu Sun.

Figure 1
Figure 1. Figure 1: Deployment scenario. Stage 1, wearable: the pulse is monitored on its own, and the PPG agent reads it as far as the pulse allows—an irregular rhythm it calls atrial fib￾rillation, flagged as needing an ECG to confirm. Stage 2, monitored: the ECG is added, and Cardiologent reads it together with the pulse and the patient’s own data to reach the decision—the therapy indicated, and how urgent. Ex￾cerpts are v… view at source ↗
Figure 2
Figure 2. Figure 2: Patient-level arrhythmia decision support on a monitored record. An 80-year-old man with hypertension and diabetes, whose record holds a burden of premature ventricular contractions (PVC). Every system gets the same input: his metadata and the windowed lead II ECG and PPG. Blue marks the clinical response, pink the evidence for it. GPT-4o prescribes rate control and anticoagulation with no evidence at all,… view at source ↗
Figure 3
Figure 3. Figure 3: Cardiologent. Window level: an agent per signal reads the waveform through its tools—seven on the ECG, nine on the PPG—and returns a grounded class, and window fusion reconciles the ECG and PPG readings by debate into one verdict. Patient level: the verdicts are aggregated into a rhythm profile, reconciled across signals by the same debate, and only then read against the patient’s metadata through guidelin… view at source ↗
Figure 4
Figure 4. Figure 4: Rater agreement (ICC(2,1), absolute agreement, with 95% CI) on the 20-patient subset. The judge’s agree￾ment with each cardiologist (0.74, 0.66) is comparable to the two cardiologists’ agreement with each other (0.67): the judge agrees with a cardiologist about as closely as the two cardiologists agree with each other. a wearable: our pulse comes from a monitor, on a still patient with a fixed sensor. The … view at source ↗
Figure 5
Figure 5. Figure 5: The specialist. Stage 1 pretrains a time-series encoder by masked reconstruction of the raw waveform. Stage 2 is the [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Grounding on ECG window 4925_w0099 (ground-truth PAC). Each row is one model’s localized ab￾normal span on the same trace; the top row is the ground truth. Cardiologent localizes a tight span on the PAC, whereas the baselines mark almost the entire window, inflating IoU while collapsing MCC toward zero. from. It is not a uniform edge on the common rhythms: it is that the general baselines collapse to zero … view at source ↗
Figure 7
Figure 7. Figure 7: Case study: Window fusion (Case2863_W0000; GT: atrial fibrillation). The ECG agent misreads an ambigu￾ous P wave and calls MAT; the PPG agent calls AF from the R–R irregularity. Because the two conflict, fusion runs a two￾round analyst–devil’s-advocate debate: the devil’s advocate notes the PPG cannot confirm P waves, the analysts weigh this against the measured irregularity (R–R CV 0.274/0.234), and the r… view at source ↗
Figure 8
Figure 8. Figure 8: Case study: Patient-level (case 2218; GT: atrial fibrillation). The ECG and PPG readings agree, so patient fu￾sion returns AF as the dominant rhythm without a conflict to settle. Retrieval supplies the guideline that sets how strongly the recommendation may be put—CHA2DS2-VASc 1 in a man is intermediate, and anticoagulation is licensed by doc￾umented AF rather than by the score—and the critic removes the d… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

54 extracted references · 7 linked inside Pith

  1. [1]

    Nature communications , volume=

    Automatic diagnosis of the 12-lead ECG using a deep neural network , author=. Nature communications , volume=. 2020 , publisher=

    2020

  2. [2]

    Journal of the American College of Cardiology , volume=

    2017 AHA/ACC/HRS guideline for management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society , author=. Journal of the American College of Cardiology , volume=. 2018 , publisher=

    2017

  3. [3]

    arXiv preprint arXiv:2405.14755 , year=

    Large language models can be zero-shot anomaly detectors for time series? , author=. arXiv preprint arXiv:2405.14755 , year=

    Pith/arXiv arXiv

  4. [4]

    2025 , eprint=

    Qwen2.5 Technical Report , author=. 2025 , eprint=

    2025

  5. [5]

    BioData mining , volume=

    The Matthews correlation coefficient (MCC) is more reliable than balanced accuracy, bookmaker informedness, and markedness in two-class confusion matrix evaluation , author=. BioData mining , volume=. 2021 , publisher=

    2021

  6. [6]

    arXiv preprint arXiv:2601.20323 , year=

    ECG-Agent: On-Device Tool-Calling Agent for ECG Multi-Turn Dialogue , author=. arXiv preprint arXiv:2601.20323 , year=

    arXiv

  7. [7]

    arXiv preprint arXiv:2507.06261 , year=

    Gemini 2.5: Pushing the frontier with advanced reasoning, multimodality, long context, and next generation agentic capabilities , author=. arXiv preprint arXiv:2507.06261 , year=

    Pith/arXiv arXiv

  8. [8]

    arXiv preprint arXiv:2305.14325 , year=

    Improving factuality and reasoning in language models through multiagent debate , author=. arXiv preprint arXiv:2305.14325 , year=

    Pith/arXiv arXiv

  9. [9]

    Scientific Data , volume=

    Vitaldb arrhythmia database: An anesthesiologist-validated large-scale intraoperative arrhythmia dataset with beat and rhythm labels , author=. Scientific Data , volume=. 2026 , publisher=

    2026

  10. [10]

    Journal of internal medicine , volume=

    Asymptomatic versus symptomatic persistent atrial fibrillation: clinical and noninvasive characteristics , author=. Journal of internal medicine , volume=. 2001 , publisher=

    2001

  11. [11]

    Cell biochemistry and biophysics , volume=

    Cardiac arrhythmias: diagnosis, symptoms, and treatments , author=. Cell biochemistry and biophysics , volume=. 2015 , publisher=

    2015

  12. [12]

    Journal of the American College of Cardiology , volume=

    Mobile photoplethysmographic technology to detect atrial fibrillation , author=. Journal of the American College of Cardiology , volume=. 2019 , publisher=

    2019

  13. [13]

    Sensors , volume=

    Premature atrial and ventricular contraction detection using photoplethysmographic data from a smartwatch , author=. Sensors , volume=. 2020 , publisher=

    2020

  14. [14]

    Nature medicine , volume=

    Cardiologist-level arrhythmia detection and classification in ambulatory electrocardiograms using a deep neural network , author=. Nature medicine , volume=. 2019 , publisher=

    2019

  15. [15]

    Computers in biology and medicine , volume=

    Opportunities and challenges of deep learning methods for electrocardiogram data: A systematic review , author=. Computers in biology and medicine , volume=. 2020 , publisher=

    2020

  16. [16]

    arXiv preprint arXiv:2410.21276 , year=

    Gpt-4o system card , author=. arXiv preprint arXiv:2410.21276 , year=

    Pith/arXiv arXiv

  17. [17]

    Scientific Data , volume=

    VitalDB, a high-fidelity multi-parameter vital signs database in surgical patients , author=. Scientific Data , volume=. 2022 , publisher=

    2022

  18. [18]

    Journal of the American Heart Association , volume=

    Multiclass arrhythmia detection and classification from photoplethysmography signals using a deep convolutional neural network , author=. Journal of the American Heart Association , volume=

  19. [19]

    npj Digital Medicine , volume=

    Teaching multimodal LLMs to comprehend 12-lead electrocardiographic images , author=. npj Digital Medicine , volume=. 2026 , publisher=

    2026

  20. [20]

    Computer Methods and Programs in Biomedicine , volume=

    Machine-learning guided differentiation between photoplethysmography waveforms of supraventricular and ventricular origin , author=. Computer Methods and Programs in Biomedicine , volume=. 2025 , publisher=

    2025

  21. [21]

    Biochimica et Biophysica Acta (BBA)-Protein Structure , volume=

    Comparison of the predicted and observed secondary structure of T4 phage lysozyme , author=. Biochimica et Biophysica Acta (BBA)-Protein Structure , volume=. 1975 , publisher=

    1975

  22. [22]

    NPJ digital medicine , volume=

    Photoplethysmography based atrial fibrillation detection: a review , author=. NPJ digital medicine , volume=. 2020 , publisher=

    2020

  23. [23]

    New England Journal of Medicine , volume=

    Large-scale assessment of a smartwatch to identify atrial fibrillation , author=. New England Journal of Medicine , volume=. 2019 , publisher=

    2019

  24. [24]

    2026 , eprint=

    MedGemma Technical Report , author=. 2026 , eprint=

    2026

  25. [25]

    Nature Reviews Cardiology , volume=

    Artificial intelligence-enhanced electrocardiography in cardiovascular disease management , author=. Nature Reviews Cardiology , volume=. 2021 , publisher=

    2021

  26. [26]

    TechRxiv , volume =

    Xiaoda Wang and Ching Chang and Defu Cao and Kaiqiao Han and Fang Sun and Yue Huang and Minxiao Wang and Chang Xu and Xiao Luo and Runze Yan and Xiangliang Zhang and Xiao Hu and Yan Liu and Yizhou Sun and Wei Wang and Carl Yang , title =. TechRxiv , volume =

  27. [27]

    European heart journal , volume=

    2022 ESC Guidelines for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death: Developed by the task force for the management of patients with ventricular arrhythmias and the prevention of sudden cardiac death of the European Society of Cardiology (ESC) Endorsed by the Association for European Paediatric and Co...

    2022

  28. [28]

    Fallahpour, Adibvafa and Ma, Jun and Munim, Alif and Lyu, Hongwei and Wang, Bo , booktitle =

  29. [29]

    2025 , eprint=

    A Survey on LLM-as-a-Judge , author=. 2025 , eprint=

    2025

  30. [30]

    Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

    Masked autoencoders are scalable vision learners , author=. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages=

  31. [31]

    International conference on learning representations , volume=

    Time-llm: Time series forecasting by reprogramming large language models , author=. International conference on learning representations , volume=

  32. [32]

    2026 , eprint=

    UniECG: Understanding and Generating ECG in One Unified Model , author=. 2026 , eprint=

    2026

  33. [33]

    arXiv preprint arXiv:2602.04279 , year=

    ECG-R1: Protocol-Guided and Modality-Agnostic MLLM for Reliable ECG Interpretation , author=. arXiv preprint arXiv:2602.04279 , year=

    Pith/arXiv arXiv

  34. [34]

    Circulation , volume=

    2023 ACC/AHA/ACCP/HRS guideline for the diagnosis and management of atrial fibrillation: a report of the American College of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines , author=. Circulation , volume=. 2024 , publisher=

    2023

  35. [35]

    Findings of the Association for Computational Linguistics: ACL 2024 , pages=

    DEBATE: Devil’s advocate-based assessment and text evaluation , author=. Findings of the Association for Computational Linguistics: ACL 2024 , pages=

    2024

  36. [36]

    2024 , eprint=

    MDAgents: An Adaptive Collaboration of LLMs for Medical Decision-Making , author=. 2024 , eprint=

    2024

  37. [37]

    Journal of the American College of Cardiology , volume=

    2018 ACC/AHA/HRS guideline on the evaluation and management of patients with bradycardia and cardiac conduction delay: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society , author=. Journal of the American College of Cardiology , volume=. 2019 , publisher=

    2018

  38. [38]

    Advances in Neural Information Processing Systems , volume=

    Gem: Empowering mllm for grounded ecg understanding with time series and images , author=. Advances in Neural Information Processing Systems , volume=

  39. [39]

    2024 , eprint=

    MMedAgent: Learning to Use Medical Tools with Multi-modal Agent , author=. 2024 , eprint=

    2024

  40. [40]

    Nejm ai , volume=

    An electrocardiogram foundation model built on over 10 million recordings , author=. Nejm ai , volume=. 2025 , publisher=

    2025

  41. [41]

    2025 , eprint=

    anyECG-chat: A Generalist ECG-MLLM for Flexible ECG Input and Multi-Task Understanding , author=. 2025 , eprint=

    2025

  42. [42]

    Proceedings of the 2024 conference on empirical methods in natural language processing , pages=

    Encouraging divergent thinking in large language models through multi-agent debate , author=. Proceedings of the 2024 conference on empirical methods in natural language processing , pages=

    2024

  43. [43]

    2023 , eprint=

    G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment , author=. 2023 , eprint=

    2023

  44. [44]

    Advances in neural information processing systems , volume=

    Visual instruction tuning , author=. Advances in neural information processing systems , volume=

  45. [45]

    Journal of the American College of Cardiology , volume=

    2015 ACC/AHA/HRS guideline for the management of adult patients with supraventricular tachycardia: a report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society , author=. Journal of the American College of Cardiology , volume=. 2016 , publisher=

    2015

  46. [46]

    Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Chat-ts: Enhancing multi-modal reasoning over time-series and natural language data , author=. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  47. [47]

    Findings of the association for computational linguistics: ACL 2025 , pages=

    MEIT: Multimodal electrocardiogram instruction tuning on large language models for report generation , author=. Findings of the association for computational linguistics: ACL 2025 , pages=

    2025

  48. [48]

    Health Data Science , volume=

    ECG-LM: understanding electrocardiogram with a large language model , author=. Health Data Science , volume=. 2025 , publisher=

    2025

  49. [49]

    arXiv preprint arXiv:2508.13256 , year=

    Cardaic-agents: A multimodal framework with hierarchical adaptation for cardiac care support , author=. arXiv preprint arXiv:2508.13256 , year=

    arXiv

  50. [50]

    2025 IEEE International Conference on Multimedia and Expo (ICME) , pages=

    Ecg-chat: A large ecg-language model for cardiac disease diagnosis , author=. 2025 IEEE International Conference on Multimedia and Expo (ICME) , pages=. 2025 , organization=

    2025

  51. [51]

    2023 , eprint=

    Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena , author=. 2023 , eprint=

    2023

  52. [52]

    Advances in neural information processing systems , volume=

    One fits all: Power general time series analysis by pretrained lm , author=. Advances in neural information processing systems , volume=

  53. [53]

    arXiv preprint arXiv:2410.02026 , year=

    Zodiac: A cardiologist-level llm framework for multi-agent diagnostics , author=. arXiv preprint arXiv:2410.02026 , year=

    Pith/arXiv arXiv

  54. [54]

    arXiv preprint arXiv:2410.05440 , year=

    Can llms understand time series anomalies? , author=. arXiv preprint arXiv:2410.05440 , year=

    Pith/arXiv arXiv