Humasak Simanjuntak

—

Unclaimed · AI-structured profileOpenAlex

Applies LoRA-tuned Stable Diffusion XL with multimodal conditioning to culturally faithful Ulos motif generation

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-20) 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—

Insufficient data: no publication counts, venue history, or yearly totals were provided, and the catalog holds a single preprint. This axis would measure the volume, format, and consistency of research output over time.

Impact—

Insufficient data: no citation counts, h-index, or adoption signals exist for the single cataloged preprint. This axis would track citations and uptake of the work by other researchers and practitioners.

Focus100/100

Focus scores 100 because the raw field distribution contains exactly one entry — 'applied generative AI for cultural heritage (computer vision)', weight 1 — so all cataloged output sits in a single field. That concentration is total but rests on one 2026 first-author preprint; the perfect score should be read as a one-paper sample rather than a demonstrated long-term specialization, and future work could broaden it quickly.

Collaboration32/100

Collaboration scores 32, computed from a raw count of 4 distinct coauthors — Tamara Yunika Sianipar, Bronson T. M Siallagan, Difya Laurensya Ambarita, and Arlinta Barus — all of whom appear on the single 2026 preprint, where Simanjuntak is first author but not corresponding author. The moderate score reflects a small, single-project team with no visible cross-institutional or international spread; the network cannot be characterized beyond these four names.

Activity—

Insufficient data: the record contains only one 2026 paper and no year-by-year publication series. This axis would measure publishing cadence and sustained research engagement across recent years.

Reproducibility20/100

Reproducibility scores 20 — the mean over exactly 1 assessed paper, which Evolve rates low readiness. The review names concrete barriers: a 7-image test set, inconsistent metric reporting, and no text-only baseline; the catalog data records no code or model release either. Because the sample is a single preprint, this is an early, weak signal about practice rather than a settled pattern, and it could change with one carefully documented follow-up.

Research summary

Humasak Simanjuntak is an early-stage researcher working at the intersection of generative AI and cultural-heritage computer vision; this profile is built from public bibliographic data. The Evolve catalog records one paper, a 2026 arXiv preprint with Simanjuntak as first author: 'Multimodal Conditioning of Fine-Tuned Stable Diffusion XL for Controllable and Culturally Faithful Ulos Motif Generation.' The work LoRA fine-tunes Stable Diffusion XL and combines LLaVA and ControlNet conditioning to steer generation toward Ulos, a traditional Indonesian woven-textile motif vocabulary. Its most interesting reported finding, per Evolve's review, is an ablation showing that stacking more conditioning signals does not improve results — a contrarian data point against the usual more-control-is-better assumption. Evolve's verdict tempers the enthusiasm: evidence rests on a seven-image test set, metric reporting is inconsistent, and no text-only baseline is included, so reproducibility readiness is rated low (reproducibility dimension 20/100, computed from one paper). The focus score is 100 — every cataloged output sits in a single field — while collaboration scores 32 on four distinct coauthors (Tamara Yunika Sianipar, Bronson T. M Siallagan, Difya Laurensya Ambarita, Arlinta Barus). Output, impact, and activity dimensions are missing for insufficient data, and no institution, country, or ORCID is attached, so identity resolution rests on bibliographic matching alone.

Collaboration pattern

All recorded collaboration sits on the single 2026 Ulos preprint, where Simanjuntak is first author (not corresponding) with four coauthors: Tamara Yunika Sianipar, Bronson T. M Siallagan, Difya Laurensya Ambarita, and Arlinta Barus. The collaboration dimension scores 32 on this small base. Beyond these four names, no cross-institutional or international links are visible in the data, so the wider network remains uncharacterized.

Research keywords

latent diffusion modelsStable Diffusion XLLoRA fine-tuningmultimodal conditioningControlNetLLaVAcultural heritage computingUlos textile motifstext-to-image generationcontrollable generationparameter-efficient fine-tuningablation evaluation

Highlights

LoRA-tuned SDXL with LLaVA and ControlNet conditioning for Ulos motifs
In a 2026 first-author arXiv preprint, Simanjuntak fine-tunes Stable Diffusion XL with LoRA and combines LLaVA and ControlNet conditioning to make Ulos woven-textile motif generation controllable and culturally faithful, per the paper's title and Evolve's review.
Ablation finding: more conditioning is not automatically better
Evolve's review flags the paper's ablation result — that adding more conditioning signals does not improve generation — as its most interesting finding, a contrarian data point against the default assumption that stacking control mechanisms helps.
Thin evidence base keeps reproducibility low
The verdict rests on a 7-image test set with inconsistent metric reporting and no text-only baseline; reproducibility readiness is rated low, giving a reproducibility dimension score of 20/100 from one assessed paper.

Timeline

2026First-author arXiv preprint on multimodal conditioning of fine-tuned Stable Diffusion XL for Ulos motif generation appears in the Evolve catalog.
2026The paper reports an ablation finding that adding more conditioning signals does not improve generation quality.
2026Evolve reviews the paper, rating reproducibility readiness low (20/100) and citing a 7-image test set, inconsistent metrics, and no text-only baseline.

Papers in the Evolve catalog

2026-09-17first author
Multimodal Conditioning of Fine-Tuned Stable Diffusion XL for Controllable and Culturally Faithful Ulos Motif Generation

Frequent co-authors

Tamara Yunika Sianipar×1 Bronson T. M Siallagan×1 Difya Laurensya Ambarita×1 Arlinta Barus×1

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.

All dimension scores rest on a one-paper sample (a single 2026 arXiv preprint); output, impact, and activity axes are missing for insufficient data.
No institution, country, or ORCID is attached; identity is name-resolved from public bibliographic data and may be ambiguous.
The reproducibility score (20) reflects Evolve's review of one paper rated low readiness — an early signal, not a settled pattern.
No citation counts, venue acceptance, or code/model release information was provided; nothing here should be read as an impact endorsement.