Seungyoon Choi
Korea Advanced Institute of Science and Technology · South Korea
KAIST researcher linking graph neural networks for credit risk with LLM-based online time series forecasting
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
This axis measures the volume and consistency of publication output over time. The current snapshot contains no scored output history — only a 2026 count of three works — so a productivity trajectory cannot yet be established and the axis is left unscored.
Impact—
Impact measures citation-based influence such as accumulated citations, h-index, and i10-index. The snapshot records both indicators at zero with no citation accumulation yet, so this axis cannot be scored at this early stage.
Focus0/100
The focus score is 0, reflecting an even split (0.5 and 0.5) between two distinct fields: machine learning for finance (credit risk) and machine learning for time-series forecasting. Topic records span five clusters — advanced graph neural networks (2), complex network analysis, stock-market forecasting, credit risk and financial regulation, and financial distress and bankruptcy prediction — indicating breadth across two research lines rather than concentration in one.
Collaboration46/100
The collaboration score is 46, based on nine distinct coauthors recorded in the snapshot. Recurring partners include Junghoon Kim, Hyunsung Kim, KyoungYong Park, Jihun Lee, YongGu Ji, Chanyoung Park, Hyunchul Kim, and Jae-Gil Lee. Both reviewed papers are multi-authored — Choi leads one as first author and sits mid-list on the other — and neither records a corresponding-author role, consistent with a moderate, team-embedded collaboration pattern.
Activity40/100
The activity score is 40. The snapshot's counts-by-year records three works in 2026, and recent three-year works also total three — meaning all dated output falls within a single year with no longer publication history to smooth it. This front-loaded, recently started record yields a moderate activity rating that should stabilize as more years accumulate.
Reproducibility60/100
The reproducibility score is 60, a mean taken across the two papers Evolve has assessed. The first-authored CoSPOT paper is rated high readiness, while DefaultGNN is rated low because its evidence rests on a proprietary Korean tax-invoice dataset. The split suggests reproducibility here depends heavily on data accessibility rather than method transparency alone.
Research summary
Seungyoon Choi is an early-stage researcher at the Korea Advanced Institute of Science and Technology (KAIST) whose recorded work, per OpenAlex bibliographic data, sits across two machine-learning areas: credit-risk modeling for finance and time-series forecasting. The snapshot records three works in 2026 — the entire dated output on file — while citation indicators (h-index and i10-index) currently stand at zero, consistent with a record that has only just begun to accumulate. Evolve has reviewed two 2026 arXiv papers. The first-authored CoSPOT paper proposes compositional spectral prompts that adapt a frozen large language model to online time-series forecasting, reporting strong gains; Evolve rates its reproducibility readiness high while flagging thin significance reporting and heavy offline cost. The second, DefaultGNN, a middle-authorship contribution, applies a dual buyer/seller-perspective graph neural network to corporate default prediction on a large Korean tax-invoice network; Evolve judges the study solid but proprietary-data-bound with modest gains, and rates reproducibility readiness low. Across nine distinct coauthors, Choi works within Korean machine-learning teams, holding first authorship in one line and a middle position in the other. The profile splits evenly (0.5 and 0.5) across the two fields, and overall dimensions remain provisional pending citation accumulation.
Collaboration pattern
Choi's recorded network comprises nine distinct coauthors, centered on Korean machine-learning groups. Recurring partners include Junghoon Kim, Chanyoung Park, Jae-Gil Lee, Hyunsung Kim, KyoungYong Park, Jihun Lee, YongGu Ji, and Hyunchul Kim. The pattern is team-based rather than solo: Choi holds first authorship on the forecasting paper and a middle position on the credit-risk paper, with no corresponding-author role recorded in either.
Research keywords
Highlights
CoSPOT: frozen-LLM spectral prompts for online time-series forecasting (first author)
DefaultGNN: dual-view GNN for corporate default from buyer-seller networks (co-author)
Timeline
- 2026Posts CoSPOT to arXiv as first author, proposing compositional spectral prompts for LLM-based online time-series forecasting.
- 2026Co-authors DefaultGNN on arXiv, a dual-perspective GNN for corporate default prediction from buyer-seller transaction networks.
- 2026OpenAlex records three works for the year — the entire dated output in the snapshot — while h-index and i10-index remain at zero.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometrics are drawn from OpenAlex only; no ORCID is linked, so this profile is name-resolved from public bibliographic data and 'Seungyoon Choi' is a common Korean name.
- The output and impact dimensions are not scored: the snapshot records h-index 0 and i10-index 0, with citation accumulation yet to begin.
- Evolve has reviewed 2 of the 3 works recorded for 2026; the remaining work is outside the reviewed catalog and is not characterized here.
- Both catalog papers are 2026 arXiv preprints; peer-reviewed venue status is not established in the available data.