NUS speech-and-dialogue researcher now shaping LLM post-training, moral-agency evaluation, and human-AI collaboration.
5
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
26
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—
A full output score is not available yet: this snapshot provides recent three-year work counts but no aggregated lifetime publication total or venue-type breakdown. This axis would measure cumulative scholarly production and its trajectory across the whole career.
Impact—
An impact score has not been computed for this snapshot: citation-side aggregation for the axis is missing, even though raw h-index (26) and i10-index (72) are reported. It would measure citation influence, uptake, and scholarly reference flow to this body of work.
Focus0/100
The focus score of 0 reflects deliberately broad coverage rather than a narrow specialty: the field distribution splits almost evenly (~33% each) across NLP/computational moral psychology, ML/NLP LLM post-training, and human-AI collaboration, while top OpenAlex topics span speech recognition and synthesis (118), topic modeling (105), speech and dialogue systems (78), speech and audio processing (72), and NLP techniques (66).
Collaboration84/100
Scoring 84, collaboration is a clear strength: 66 distinct coauthors appear in the recent three-year window alone, and the frequent-coauthor list spans long-standing NUS colleagues (Zhengyuan Liu, Vernon Toh, Aloysius Tok, Jingtan Wang), the affective-computing cluster around Soujanya Poria, Navonil Majumder and Fernanda Mansilla, and international partners (Farah Benamara, Bahia Guellaï, Daniela Rus, Arun Verma, Xiaoqiang Lin).
Activity100/100
The maximum activity score of 100 matches the record: 72 works in the recent three-year window, with annual counts climbing from 19 (2022) to 31 (2023), 29 (2024), and 34 (2025), and 9 works already logged for 2026. The catalog adds five arXiv papers that year — an unusually dense burst across post-training, agent benchmarks, and human-AI collaboration.
Reproducibility58/100
Reproducibility averages 58 across the 3 papers Evolve has assessed. The spread is wide: the SFT–RL budget-allocation study and the MNIST-PRO benchmark rate high, the moral-agency study rates medium (conclusions rest on 4 small models), and the workflow-level augmentation framework rates low as a conceptually solid but empirically untested whitepaper — strong on empirical work, weaker on position-style frameworks.
Research summary
Nancy F. Chen is a researcher at the National University of Singapore whose profile is name-resolved from public OpenAlex bibliographic data. The record shows a long, steadily compounding publication history — h-index 26 and i10-index 72, with works recorded back to 2007 — that has accelerated sharply: 72 works in the recent three-year window, peaking at 34 in 2025, yielding a maximum activity score of 100. Her topical roots lie in speech: OpenAlex topics are led by speech recognition and synthesis (118), topic modeling (105), speech and dialogue systems (78), speech and audio processing (72), and NLP techniques (66). Her current Evolve-cataloged work pivots to LLM-era questions across three evenly weighted fields: NLP/computational moral psychology, ML/NLP LLM post-training, and human-AI collaboration and the future of work. Five 2026 arXiv papers appear in the catalog: a credible SFT–RL annotation-budget allocation study, the MNIST-PRO partially observable agent benchmark, the first human-vs-LLM study of perceived moral agency (last-authored), a workflow-level framework for genuine AI augmentation, and the multilingual CultureConverse harness. Reproducibility averages 58 across three assessed papers — strong on empirical benchmarks, weaker on conceptual frameworks. Collaboration is a clear strength (84), with 66 distinct recent coauthors spanning NUS, the affective-computing community, and international partners.
Collaboration pattern
Collaboration scores 84, backed by 66 distinct coauthors in the recent three-year window. The frequent-coauthor list mixes long-standing NUS colleagues (Zhengyuan Liu, Vernon Toh, Aloysius Tok, Jingtan Wang), the affective-computing cluster around Soujanya Poria, Navonil Majumder and Fernanda Mansilla, international partners (Farah Benamara, Bahia Guellaï, Daniela Rus), and newer collaborators (Arun Verma, Xiaoqiang Lin). Recent catalog papers place her in middle and last authorship, consistent with senior, team-based research.
Research keywords
Highlights
Near-Optimal SFT–RL Budget Allocation Scales Across LLM Sizes
MNIST-PRO Recasts MNIST as a Partially Observable Agent World
First Human-vs-LLM Study of Perceived Moral Agency
A Workflow-Level Framework for When AI Truly Augments Work
Timeline
- 2007Earliest year in the OpenAlex counts-by-year record, with 2 works — the starting point of a publication history that continues unbroken through 2026.
- 2016Output reaches 22 works, the highest annual count of the pre-2021 record, coinciding with the strongest speech-and-dialogue topical cluster.
- 202124 works as the recent acceleration begins; the annual record then rises through 19 (2022) to 31 (2023).
- 202534 works, the highest annual count in the provided record, within a 72-work recent three-year window that earns a maximum activity score of 100.
- 2026Five arXiv papers enter the Evolve catalog: the SFT–RL budget-allocation study, the MNIST-PRO benchmark, the moral-agency evaluation, the workflow-level augmentation framework, and the CultureConverse harness.
Papers in the Evolve catalog
When Does AI Augment Work? A Workflow-Level Framework for Human-Agent Collaboration
How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions
Scaling Near-Optimal SFT-RL Annotation Budget Allocation from Small to Large LLMs
MNIST-PRO: MNIST is Back as a Partially Observable World for AI Agents
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 figures are drawn from a single OpenAlex snapshot; h-index 26 and i10-index 72 reflect that source only.
- Output and impact radar axes are unscored because the underlying lifetime aggregation data is insufficient in this snapshot.
- CultureConverse is cataloged but not yet reviewed by Evolve — no verdict, keywords, or reproducibility rating exist for it.
- All five catalog papers are 2026 arXiv preprints; peer-review venue status is not captured in the provided data.