ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution

Xiaoliang Wang; Zhuo Chen; Chuanyang Zheng; Jiangwei Lao; Quanbin Wang; Honglin Lin; Jiajia Liu; Zhihong Lu

cs.AIOct 13, 2026

ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution

Authors: Xiaoliang Wang, Zhuo Chen, Chuanyang Zheng, Jiangwei Lao, Quanbin Wang, Honglin Lin, Jiajia Liu, Zhihong Lu,

Organizations: Zhejiang University · Ant Group · Shanghai Artificial Intelligence Laboratory · Shanghai Jiao Tong University

Abstract

Electrocardiogram (ECG) models need supervision that targets evolving weaknesses and remains faithful to waveform evidence. We introduce ECG-RSI, a recursive self-improvement framework combining verified waveform synthesis with feedback-guided intervention search. Signal remeasurement verifies waveform edits; construction errors guide generator repair, while learner errors motivate interventions compared on a separate selection set. Across seven benchmarks, three rounds raise Qwen3.5-35B-A3B's mean question-answering accuracy from 34.57% to 56.04% and mean classification F1 from 24.27 to 34.03. Across the retained trajectory, the preferred supervision shifts from direct reinforcement to confusion-focused contrasts. Continued feedback correction yields diminishing external gains. These results motivate separating error diagnosis from intervention selection and checking whether gains on the feedback set transfer to independent benchmarks.

Figures & tables

Figure 1: ECG-RSI : feedback-driven data and model evolution. The data factory constructs and verifies ECG questions, with rejected candidates driving local revisions of operators, prompts, and skills. The main agent diagnoses feedback, proposes interventions, and selects accepted data for training. The training workshop evaluates candidate checkpoints and supplies category-, task-, and difficulty-level feedback. Independent evaluation is separated from the update loop. The search tree and highlighted path illustrate candidate comparison and model selection.

Figure 2: Controllable waveform perturbations across rhythm, intervals, morphology, amplitude, and individual leads. The examples illustrate how an intended feature change is translated into a localized signal edit. Edited waveforms are remeasured to check the target and protected features before question generation, linking controllable signal changes to evidence-grounded supervision .

Figure 3: Schematic of feedback-guided model evolution. A/B/C change category coverage, confusion contrasts, or difficulty/task mixtures. Each candidate adds 10,000 accepted questions plus replay. Purple shows an illustrative lineage. Tab. 12 reports measured mixtures, scores, and choices.

Panel of Table 4 — Tab. 1 : Comparison of ECG-RSI with proprietary, open-source, and ECG-specific models on seven benchmarks. The first four columns report question-answering accuracy and the next three report classification F1; S7 is their equal-weight mean on a 0–100 scale. ECG-RSI R3 achieves the highest aggregate score.

R-A round	MMMU ECG Acc.	CSN Acc.	G12EC Acc.	ECG-QA Acc.	PTB-XL F1	CPSC 2018 F1	CODE 15% F1	S7
R0	42.00	30.91	41.36	23.99	29.50	13.10	30.20	30.15
R1	48.00 $\uparrow$	61.68 $\uparrow$	64.61 $\uparrow$	33.56 $\uparrow$	32.90 $\uparrow$	23.00 $\uparrow$	38.80 $\uparrow$	43.22 $\uparrow$
R2	56.00 $\uparrow$	64.86 $\uparrow$	66.39 $\uparrow$	34.55 $\uparrow$	36.40 $\uparrow$	23.90 $\uparrow$	41.50 $\uparrow$	46.23 $\uparrow$
R3	57.50 $\uparrow$	67.23 $\uparrow$	66.24 $\downarrow$	33.18 $\downarrow$	38.30 $\uparrow$	22.80 $\downarrow$	41.00 $\downarrow$	46.61 $\uparrow$
$\Delta$ (R3 $-$ R0)	+15.50 $\uparrow$	+36.32 $\uparrow$	+24.88 $\uparrow$	+9.19 $\uparrow$	+8.80 $\uparrow$	+9.70 $\uparrow$	+10.80 $\uparrow$	+16.46 $\uparrow$

Tab. 2 : Round-wise benchmark performance of series R-A. R0 is the starting model; each accepted update adds 10,000 new questions while retaining historical training data. Bold and underlining mark the best and second-best values within each column; blue/red arrows indicate increases/decreases from the previous round, and blue shading marks R3. Aggregate performance improves through R3, although several individual metrics peak at R2.

	Accuracy (%)				Accuracy difference
Dataset	R0	R1	R2	R3	R1 $-$ R0	R2 $-$ R1	R3 $-$ R2	R3 $-$ R0
MedX	51.57	55.94	55.86	57.24	+4.37	-0.08	+1.38	+5.67
LUDB	53.45	61.28	63.28	67.24	+7.83	+2.00	+3.96	+13.79

Tab. 3 : Validation accuracy on MedX and LUDB across four rounds. The change columns report later minus earlier accuracy, in percentage points. Both datasets improve overall; LUDB increases at every round, whereas MedX shows a small intermediate decline .

Figure 4: Round-wise benchmark, validation, and strategy results. (a) Four multiple-choice question (MCQ) accuracy metrics; (b) three classification F1 metrics; (c) MedX and LUDB validation accuracy; (d) seven-benchmark aggregate score (S7) for tree, uniform-allocation, and chain trajectories. Higher values indicate better performance. Error bars show

\pm 1

standard deviation (SD) where available; LUDB has no SD bars. The panels distinguish broad gains from task-specific fluctuations and the late-round declines of the two strategy controls.

Round	NSR	AF	AFL	ST	SB	PVC	PAC	AVB-I	Mobitz-I	Mobitz-II	AVB-III	LBBB	RBBB	Ant-STEMI	Inf-STEMI	Lat-STEMI	Post-STEMI	LVH	RVH	LQTS	HyperK	WPW	Brugada
$N$	168	154	147	132	128	143	136	121	109	105	118	139	141	149	145	116	108	152	124	134	117	113	101
R0	25.0	39.6	44.9	28.0	26.6	40.6	39.0	38.0	42.2	48.6	44.1	36.7	35.5	44.3	43.4	46.6	49.1	37.5	41.1	41.0	42.7	41.6	45.5
R1	19.6 $\scriptstyle\downarrow$	32.5 $\scriptstyle\downarrow$	38.1 $\scriptstyle\downarrow$	22.0 $\scriptstyle\downarrow$	20.3 $\scriptstyle\downarrow$	33.6 $\scriptstyle\downarrow$	32.4 $\scriptstyle\downarrow$	31.4 $\scriptstyle\downarrow$	34.9 $\scriptstyle\downarrow$	41.0 $\scriptstyle\downarrow$	36.4 $\scriptstyle\downarrow$	30.2 $\scriptstyle\downarrow$	29.1 $\scriptstyle\downarrow$	37.6 $\scriptstyle\downarrow$	36.6 $\scriptstyle\downarrow$	38.8 $\scriptstyle\downarrow$	40.7 $\scriptstyle\downarrow$	30.9 $\scriptstyle\downarrow$	33.9 $\scriptstyle\downarrow$	34.3 $\scriptstyle\downarrow$	35.9 $\scriptstyle\downarrow$	34.5 $\scriptstyle\downarrow$	37.6 $\scriptstyle\downarrow$
R2	14.9 $\scriptstyle\downarrow$	26.6 $\scriptstyle\downarrow$	32.0 $\scriptstyle\downarrow$	16.7 $\scriptstyle\downarrow$	15.6 $\scriptstyle\downarrow$	27.3 $\scriptstyle\downarrow$	26.5 $\scriptstyle\downarrow$	25.6 $\scriptstyle\downarrow$	28.4 $\scriptstyle\downarrow$	34.3 $\scriptstyle\downarrow$	29.7 $\scriptstyle\downarrow$	24.5 $\scriptstyle\downarrow$	23.4 $\scriptstyle\downarrow$	31.5 $\scriptstyle\downarrow$	30.3 $\scriptstyle\downarrow$	31.9 $\scriptstyle\downarrow$	34.3 $\scriptstyle\downarrow$	25.0 $\scriptstyle\downarrow$	27.4 $\scriptstyle\downarrow$	28.4 $\scriptstyle\downarrow$	29.9 $\scriptstyle\downarrow$	28.3 $\scriptstyle\downarrow$	30.7 $\scriptstyle\downarrow$
R3	11.3 $\scriptstyle\downarrow$	21.4 $\scriptstyle\downarrow$	26.5 $\scriptstyle\downarrow$	12.9 $\scriptstyle\downarrow$	11.7 $\scriptstyle\downarrow$	21.7 $\scriptstyle\downarrow$	21.3 $\scriptstyle\downarrow$	19.8 $\scriptstyle\downarrow$	22.0 $\scriptstyle\downarrow$	27.6 $\scriptstyle\downarrow$	23.7 $\scriptstyle\downarrow$	19.4 $\scriptstyle\downarrow$	18.4 $\scriptstyle\downarrow$	26.2 $\scriptstyle\downarrow$	24.8 $\scriptstyle\downarrow$	25.9 $\scriptstyle\downarrow$	27.8 $\scriptstyle\downarrow$	19.7 $\scriptstyle\downarrow$	21.8 $\scriptstyle\downarrow$	22.4 $\scriptstyle\downarrow$	23.9 $\scriptstyle\downarrow$	22.1 $\scriptstyle\downarrow$	23.8 $\scriptstyle\downarrow$
$\Delta_{0\to 3}\!\downarrow$	13.7	18.2	18.4	15.2	14.8	18.9	17.6	18.2	20.2	21.0	20.3	17.3	17.0	18.1	18.6	20.7	21.3	17.8	19.4	18.7	18.8	19.5	21.8

Tab. 5 : ECG-category feedback error rates (%).

N

: recorded sample count;

\Delta_{0\to 3}

: R0 minus R3 error rate, in percentage points (computed before rounding). Blue marks R3; bold/underline mark the best/second-best rates;

\downarrow

marks reductions. Full names and counts: Tab. B.2 .

Feature	Unit	Bias	MAE	RMSE	$s_{e}^{2}$	MAE limit	Status
HR	bpm	0.63	2.14	2.22	4.53	$\leq 3$	Met
RR	ms	3.11	15.28	15.36	226.28	$<20$	Met
PR	ms	0.57	7.71	7.78	60.21	$\leq 25$	Met
QRS	ms	4.12	4.58	4.61	4.28	$\leq 30$	Met
QT	ms	0.94	9.71	9.85	96.15	$\leq 30$	Met
QTc	ms	0.86	10.22	10.34	106.19	$\leq 40$	Met

Tab. 6 : Waveform-feature errors from a single synthesis run of 10,000 samples and comparison with the adopted MAE criteria. Blue marks MAE values meeting the adopted criteria.

Metric	Value	95% CI	$n$	Source
Target attainment	0.82	[0.76, 0.87]	240	Stage 2
Non-target preservation	0.79	[0.73, 0.84]	240	Stage 2
Morphology consistency	0.86	[0.81, 0.90]	240	Stage 1
Lead consistency	0.83	[0.78, 0.88]	240	Stage 1
Answer accuracy	0.88	[0.83, 0.92]	240	Stage 3
Explanation support	0.76	[0.70, 0.81]	240	Stage 3

Tab. 7 : Physician review of 240 examples (95% confidence intervals). Stages 1–3 assess morphology, editing fidelity, and answers/explanations.

Edit	$n$	Target	Preserve	Morph.	Answer	Support
HR	40	0.85	0.80	0.88	0.90	0.78
RR	40	0.78	0.75	0.82	0.85	0.72
PR	40	0.80	0.78	0.85	0.88	0.75
QRS	40	0.83	0.81	0.87	0.89	0.77
QT	40	0.82	0.79	0.86	0.87	0.76
QTc	40	0.84	0.80	0.88	0.89	0.78

Tab. 8 : Clinical review by edit type: 40 examples per type, 240 overall. Values are aggregated physician ratings; higher is better.

Figure 5: Recorded scores from five models for synthetic ECG questions across 23 categories (0–100 scale). The bottom row is the per-category arithmetic mean across models ; the unweighted mean over all 115 model–category cells is 85.09, reported as the full-method Question score in Tab. 10 . The shared color scale spans 60–100, from light blue (low) to dark blue (high).

Group	Added data	S7	$\Delta$ S7
A	Ordinary training-side questions	44.15	+0.93
B	Forward-only synthesis	44.27	+1.05
C	Forward + reverse, fixed selection proportions	45.56	+2.34
D	Forward + reverse, feedback-guided selection	46.23	+3.01

Tab. 9 : Data-construction and selection ablation from a shared checkpoint with S7 = 43.22.

Method	Candidates required	Pass (%)	Question score	S7	$\Delta$ S7
Full method	16,193	61.76	85.09	43.22	0.00
Freeze operator updates	36,723	27.23	71.90	39.29	-3.93
Freeze skill updates	22,398	44.65	79.20	40.15	-3.07

Tab. 10 : R1 synthesis efficiency and training outcomes at a fixed target of 10,000 accepted questions. Candidate counts include rejected questions, and the pass rate measures the fraction accepted. Question scores assess generated-data quality, whereas S7 measures downstream model performance;

\Delta

S7 is relative to the full method. Freezing operator or skill updates increases candidate requirements and reduces both quality scores and downstream performance.

Feedback used for selection	S7	$\Delta$ S7	$\Delta$ S7 vs. fixed
None: fixed mixture	45.56	+2.34	0.00
ECG category	45.60	+2.38	+0.04
Category + question type	45.98	+2.76	+0.42
Category + type + difficulty + confusion pairs	46.23	+3.01	+0.67

Tab. 11 : Feedback granularity under a common parent and a fixed budget of 10,000 added questions per configuration.

\Delta

S7 is relative to the parent score of 43.22. The final column isolates the gain over fixed selection.

Round	Branch	New-data composition	$S_{\mathrm{sel}}$	Decision
R0 $\to$ R1	A	AV block 45%; L3 50%; measurement questions 60%	42.66	✓ Retain
	B	Mobitz I/II and LBBB/RBBB contrasts 55%; L3 40%	40.51	Archive
	C	L2+L3 60%; clinical-reasoning questions 50%	38.27	Archive
R1 $\to$ R2	A	Four regional STEMI categories 50%; L3 55%; image-localization questions 45%	44.12	Archive
	B	Anterior/lateral and inferior/posterior STEMI contrasts 60%; L3 45%	45.08	✓ Retain
	C	L3 70%; multistep-reasoning questions 55%	43.55	Archive

Tab. 12 : Actual candidate compositions and branch choices across three rounds. A, B, and C denote direct reinforcement, confusion contrast, and difficulty/task curriculum. Exactly one branch is retained per round. Percentages describe overlapping attributes of each branch’s 10,000 examples. Pair percentages cover the listed pair families jointly.

Update	Retained data	Error rate (%): before $\to$ after
R0 $\to$ R1 (A)	AV block: 4,500 (45%); L3: 5,000 (50%); measurement: 6,000 (60%)	L3: 58.20 $\to$ 51.81 ; Mobitz II: 48.57 $\to$ 40.95
R1 $\to$ R2 (B)	Regional STEMI contrasts: 6,000 (60%); L3: 4,500 (45%)	L3: 51.81 $\to$ 44.41 ; posterior STEMI: 40.74 $\to$ 34.26
R2 $\to$ R3 (B)	AF/AFL and PAC/PVC contrasts: 5,500 (55%); L3: 4,500 (45%)	L3: 44.41 $\to$ 38.17 ; AF: 26.62 $\to$ 21.43 ; AFL: 31.97 $\to$ 26.53

Tab. 13 : Data retained along the selected lineage and subsequent feedback outcomes. Counts are derived from the winning branch percentages and its 10,000 accepted questions. Category/pair counts overlap with L3 counts. Blue shading marks the final update; bold counts identify the targeted data, and blue values mark the updated error rates.

Strategy	R0	R1	R2	R3	R3 $-$ R0	R3 $-$ R2	R3 vs. tree
Tree	30.15	43.22	46.23	46.61	+16.46	+0.38	0.00
Uniform	30.15	39.67	43.25	41.45	+11.30	-1.80	-5.16
Chain	30.15	44.20	44.75	43.70	+13.55	-1.05	-2.91

Tab. 14 : S7 trajectories in R-A. The last three columns report total gain, the final-round change, and the final gap to tree. Blue shading marks tree; all differences use the S7 scale.

Method	S7 (mean $\pm$ SD)	$\Delta$
Full method	$\mathbf{43.22}\pm 1.40$	0.00
Freeze operators	$39.29\pm 1.30$	${-3.93}$ $\downarrow$
Freeze skills	$\underline{40.15}\pm 1.30$	${-3.07}$ $\downarrow$

Tab. 15 : Synthesis-update ablations and historical strategy trajectories (S7, higher is better). Left: R1 mean

\pm

standard deviation;

\Delta

is relative to the full method. Right: round-wise S7;

\Delta_{0\to 3}

is the total improvement. Blue/red arrows denote increases/decreases from the reference or previous round. Bold/underlining mark the best/second-best scores in each comparison. Blue shading highlights the full method and R3.

Appendix figures & tables6 assets

Supplementary material from the paper’s appendix.

Appendix

Round	L1 $N=1{,}218$	L2 $N=1{,}093$	L3 $N=689$
R0	318 (26.11)	421 (38.52)	401 (58.20)
R1	247 (20.28 $\downarrow$ )	362 (33.12 $\downarrow$ )	357 (51.81 $\downarrow$ )
R2	196 ( 16.09 $\downarrow$ )	309 ( 28.27 $\downarrow$ )	306 ( 44.41 $\downarrow$ )
R3	158 ( 12.97 $\downarrow$ )	267 ( 24.43 $\downarrow$ )	263 ( 38.17 $\downarrow$ )

Appendix

Tab. 16 : Difficulty-level errors: count (%). Group sizes appear in the header.

Model	Seed 1	Seed 2	Seed 3	Mean $\pm$ SD
Base	30.15	30.15	30.15	$30.15\pm 0.00$
ECG-RSI R1	33.42	33.18	33.61	$33.40\pm 0.22$
ECG-RSI R2	35.27	35.04	35.48	$35.26\pm 0.22$
ECG-RSI R3	37.86	37.52	38.09	$\bm{37.82\pm 0.29}$

Appendix

Tab. 18 : ECG-task scores across three training seeds.

Benchmark	Split	$N$	Task / reported metric
MMMU-ECG	mmmu-ecg	200	Single-choice accuracy
CSN	csn-test-no-cot	1,611	Single-choice accuracy
G12EC	g12-test-no-cot	2,026	Single-choice accuracy
ECG-QA	ecgqa-test	1,317	Answer-set exact-match accuracy
PTB-XL	ptb-test	2,082	Multi-label F1
CPSC2018	cpsc-test	2,061	Multi-label F1

Appendix

Tab. 19 : Enabled evaluation splits. The first seven contribute equally to S7; MedXpert-ECG is reported separately. Counts describe evaluation items.

Figure 6: Medical taxonomy organizing the 23 ECG task categories. Each outer sector denotes one category; inner sectors group related categories, with areas and counts indicating the number of categories in each group. The diagram summarizes task coverage and the grouping used for category-level feedback.

Figure 7: Targeted PR prolongation, shown through original and edited ECGs with lead II details. The edit lengthens the PQ segment and compensates in the TP segment to preserve the RR interval. The target PR is 300 ms and the remeasured PR is 298 ms; these automatic measurements support the subsequent parameter-comparison question.

Figure 8: Targeted heart-rate slowing, shown through original and edited ECGs with lead II details. Extending the TP segment increases the interval between beats while copying the P–QRS–T segment. The target is 48 bpm and remeasurement gives 47.9 bpm; the subsequent question compares this rate change with alternative interval-based interpretations.

Language Model Self-Improvement

Language model self-improvement studies how models use their own outputs, feedback, and experience to improve with less external supervision.

See all 171 papers

Failures Reveal What Metrics Miss: An Evidence-Driven Agent for Recursive Refinement of ECG Classifiers
Jul 27, 2026Jinliang Deng, Yiming Niu, Yibo Pan +3LLM Self-Refinement Electrocardiogram Classification
School of Computer Science and Engineering, Beihang University, Beijing, China · School of Artificial Intelligence, Beihang University, Beijing, China · School of Economics and Business Administration, Chongqing University, Chongqing, China +1
ECGQuest: Benchmarking and Fine-Tuning Language Models for Electrocardiography
Aug 31, 2026Mohammadsina Hassannia, Matthew A. Reyna, Reza SameniElectrocardiography LLM Fine-Tuning
Department of Biomedical Informatics Emory University Atlanta, GA 30322, USA · Department of Biomedical Engineering Georgia Institute of Technology and Department of Biomedical Informatics Emory University Atlanta, GA 30332, USA
The Impact of Temporal Context Length and Encoding Strategies on Self-Supervised ECG Representation Learning
Aug 13, 2026Ahmed Sameh, Ramzi Al-Sharawi, Yogatheesan VaratharajahSelf-Supervised Time Series Representation Learning Electrocardiogram Foundation Models
Computer Science & Engineering, University of Minnesota Twin Cities, Minneapolis, MN 55455. · Robotics, University of Minnesota Twin Cities, Minneapolis, MN 55455.

Jul 27, 2026J/K move · Enter open · S save

Failures Reveal What Metrics Miss: An Evidence-Driven Agent for Recursive Refinement of ECG Classifiers

Jinliang Deng, Yiming Niu, Yibo Pan, Zhiqi Shao+2

School of Computer Science and Engineering, Beihang University, Beijing, China · School of Artificial Intelligence, Beihang University, Beijing, China · School of Economics and Business Administration, Chongqing University, Chongqing, China+1

LLM Self-Refinement Electrocardiogram Classification

Open paper AskPDF arXiv

Deep models have substantially advanced 12-lead ECG classification, yet their refinement still relies heavily on human experts to inspect failures and iteratively revise classifier designs. Recent LLM-based agents have demonstrated the potential for automated model design, but when guided only by aggregate performance metrics, they lack insight into why individual cases fail and how the classifier should be revised. We present RecursiveECG, an evidence-driven LLM-as-Designer framework in which an LLM serves as an offline model designer that refines ECG classifiers based on concrete failures and objective ECG evidence. To ground failure diagnosis in executable evidence, Criteria-to-Measurement Compilation converts curated ECG criteria into validated deterministic functions that produce reproducible, reference-backed measurements for individual ECGs. Building on these measurements, Evidence-Grounded Failure Review analyzes failed and comparator cases by jointly considering raw waveforms, measurements, and model outputs, enabling the LLM to diagnose classifier limitations and formulate targeted revisions. Candidate revisions are executed and re-evaluated under a fixed problem contract, and only evidence-supported updates are retained. The resulting predictor is frozen after refinement and requires no LLM inference during deployment, while an audit trail links each accepted revision to its supporting evidence. Across PTB-XL, Georgia, and CPSC2018, RecursiveECG consistently outperforms strong baselines, achieving an average relative improvement of 10.0%. Extensive ablation and transfer studies further validate the effectiveness of its evidence-grounded refinement process.

Abstract

Language Model Self-Improvement

Explore similar work

Failures Reveal What Metrics Miss: An Evidence-Driven Agent for Recursive Refinement of ECG Classifiers

ECGQuest: Benchmarking and Fine-Tuning Language Models for Electrocardiography

The Impact of Temporal Context Length and Encoding Strategies on Self-Supervised ECG Representation Learning