Last-author NLP researcher: LLM data attribution and machine-translation evaluation, two 2026 arXiv preprints
2
papers here
0
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-27) 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—
Output cannot be scored: beyond a single 2026 entry in countsByYear, the snapshot offers no reliable publication-volume signal, and OpenAlex's h-index of 0 likely reflects incomplete name resolution. This axis would measure cumulative paper output over time.
Impact—
Impact has insufficient data: citation counts and field-weighted influence benchmarks are absent, and the recorded h-index and i10-index of 0 are unreliable here. This axis would measure citations, uptake, and influence on follow-up research.
Focus—
Focus cannot be scored: OpenAlex lists three single-count topics — Natural Language Processing Techniques, Artificial Intelligence in Healthcare and Education, AI in Service Interactions — too sparse to establish a stable research identity. This axis would measure concentration across a topic portfolio.
Collaboration36/100
Collaboration scores 36/100. The OpenAlex snapshot lists 5 distinct coauthors, and the Evolve catalog surfaces a stable core team — Sunwoo Kim, Seokwon Jung, Sohyung Kim, Seong Joon Oh, and Faiz Ghifari Haznitrama — recurring across both 2026 preprints. On each paper the researcher takes the last (senior) author position, consistent with a supervisory role in a small, tight-knit group rather than large multi-institution consortia.
Activity20/100
Activity scores 20/100. The snapshot counts only 1 work in the last three years (recent3yWorks = 1) and a single 2026 entry in OpenAlex's countsByYear, while the Evolve catalog independently holds two 2026 arXiv preprints. Both fall in the same year, so the record suggests a concentrated recent burst rather than sustained multi-year output; the underlying base is thin and any activity reading should stay cautious.
Reproducibility—
Reproducibility is not scored from artifacts; signals are mixed — the attribution paper's code is unreleased with ad hoc format classes, the MT benchmark has thorough judge validation but an admitted confound, both medium repro-readiness; it would measure artifact availability.
Research summary
Alice Oh is name-resolved here from public OpenAlex bibliographic data; the snapshot lists no institution, country, or ORCID, so biographical claims are kept to a minimum. The Evolve review catalog holds two 2026 arXiv preprints, on both of which the researcher appears in the last (senior) author position without corresponding-author status. The first, 'Form Over Content In Gradient-Based Data Attribution Methods', runs a 5×7 factorial study showing that LLM gradient attribution methods such as TracIn and LESS track answer format over task content across training regimes and model scales; Evolve finds the result convincing but notes ad hoc format classes and unreleased code. The second, 'Evaluating Communicative Success in Machine-Translated Conversation', introduces a checklist benchmark for interpreter-style machine translation with thorough LLM-judge validation, though model rankings carry an admitted context-asymmetry confound. Radar dimensions computed from real bibliometrics score collaboration at 36 and activity at 20, with 5 distinct coauthors and 1 work in the last three years; the output, impact, focus, and reproducibility axes lack sufficient data, and the recorded h-index and i10-index are both 0, likely reflecting thin name resolution for this name. A recurring co-author group — Sunwoo Kim, Seokwon Jung, Sohyung Kim, Seong Joon Oh, and Faiz Ghifari Haznitrama — suggests a compact, supervision-style collaboration pattern.
Collaboration pattern
Collaboration is compact and stable: OpenAlex records 5 distinct coauthors (radar collaboration score 36), and the same core group — Sunwoo Kim, Seokwon Jung, Sohyung Kim, Seong Joon Oh, and Faiz Ghifari Haznitrama — recurs across both 2026 preprints. Sitting in the last author slot on each paper without corresponding status suggests a senior-supervision role inside a single small team rather than distributed multi-institution partnerships; no external or industry collaborators appear in the available data.
Research keywords
Highlights
Gradient attribution tracks format over content
Checklist benchmark for interpreter-style MT
Timeline
- 2026Publishes (as last author) 'Form Over Content In Gradient-Based Data Attribution Methods', a 5×7 factorial study of TracIn- and LESS-style gradient attribution for LLMs.
- 2026Publishes (as last author) 'Evaluating Communicative Success in Machine-Translated Conversation', a checklist benchmark with LLM-judge validation for interpreter-style MT.
- 2026Both cataloged works land in the same year; OpenAlex logs 1 work in 2026 (recent3yWorks = 1) with 5 distinct coauthors.
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.
- Bibliometric scores and raw counts come from OpenAlex only; no other citation source was used.
- OpenAlex reports h-index 0 and i10-index 0 with only 1 work in 2026, which likely reflects incomplete name resolution for the name 'Alice Oh' rather than measured output; treat impact figures as unreliable.
- No institution, country, or ORCID is available, so affiliation and identity could not be verified.
- Both cataloged papers are arXiv preprints from 2026; peer-review status and final venues are not captured.