UC San Diego NLP researcher: rising output in topic modeling, dialogue personalization, and LLM scientific discovery
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
3
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—
No output score is available: the underlying per-publication venue-quality signals have not been computed for this profile. This axis would measure publication volume and venue tier over time; the 41-work total and yearly counts do not substitute for it.
Impact—
No impact score is available yet; citation data beyond h-index 3 and i10-index 1 is incomplete in this snapshot, marking an early-stage record. This axis would measure citation uptake, influence, and recognition once sufficient citation data accumulates.
Focus100/100
Focus scores 100 because the field distribution is entirely natural language processing — the only field in the snapshot — and topic assignments concentrate heavily: Topic Modeling (27) and Natural Language Processing Techniques (24) dominate, with three secondary topics at 7 each. Recent activity reinforces this, as 31 of 41 works fall in 2024–2026, though the two Evolve-reviewed 2026 preprints extend into AI-for-Science, a modest outward reach.
Collaboration70/100
Collaboration scores 70 on the strength of 33 distinct coauthors across the record and a stable recurring core of twelve frequent collaborators, including Zixuan Wang, Yufan Zhou, and Jinzhou Tang. Both Evolve-reviewed 2026 papers place Peng in a middle author position without corresponding-author status, consistent with contributing within larger team-written arXiv groups rather than leading small two-person efforts.
Activity100/100
Activity scores 100 from a steep, sustained output curve: works per year rise from 3 in 2021 and 2 in 2022 to 5 in 2023, 11 in 2024, 7 in 2025, and a record 13 in 2026. That yields 31 works in the last three years — roughly three quarters of the 41-work total — indicating a currently very high publishing cadence.
Reproducibility20/100
Reproducibility scores 20, the mean over one scored paper (reproPapers: 1, reproMean: 20), so the evidence base itself is thin. Among Evolve-reviewed work, Mind2Dialogue is rated low readiness — its Oracle teacher model is closed — while the LLM scientific-law benchmark is rated medium. No released-code or artifact information exists in the provided data beyond these ratings.
Research summary
Letian Peng is a researcher at the University of California San Diego whose public bibliographic record, resolved from OpenAlex, centers on natural language processing. The snapshot covers 41 works between 2021 and 2026, with output accelerating sharply: from 3 works in 2021 to 11 in 2024 and 13 in 2026, and 31 works in the last three years alone. Topic assignments concentrate on Topic Modeling (27) and Natural Language Processing Techniques (24), with secondary clusters in Artificial Intelligence in Games, Multimodal Machine Learning Applications, and Speech and Dialogue Systems (7 each). Two 2026 arXiv preprints sit in the Evolve catalog — Mind2Dialogue, on training human-aware assistants by simulating user mental states, and a benchmark probing whether LLMs can discover scientific laws — both with Peng in a middle author position and no corresponding-author status. Evolve's verdicts note personalization gains across three backbones in the former but flag its closed teacher model, and a well-grounded benchmark in the latter whose hidden laws are LLM-synthesized. The citation footprint is early-stage (h-index 3, i10-index 1), and reproducibility signals are thin: one paper scored at 20, with catalog readiness rated low and medium. Collaboration is broad, with 33 distinct coauthors and a recurring core of twelve.
Collaboration pattern
The record shows 33 distinct coauthors, with a recurring core of twelve frequent collaborators including Zixuan Wang, Yufan Zhou, Jinzhou Tang, Xinle Yu, and Chengjun Wu. Both Evolve-reviewed 2026 papers are middle-author, non-corresponding contributions, suggesting Peng typically works inside larger, team-written arXiv groups rather than solo or leading small authorships.
Research keywords
Highlights
Distilling an oracle into human-aware assistants (Mind2Dialogue)
Benchmarking whether LLMs can discover scientific laws
Timeline
- 2021Earliest year in the OpenAlex snapshot: 3 works recorded.
- 2022Output dips to 2 works, the lowest annual count in the record.
- 2023Annual output recovers to 5 works.
- 2024Output climbs to 11 works, the steepest rise to that point.
- 20257 works recorded, sustaining a high publishing cadence.
- 2026Most productive year with 13 works; both Evolve-reviewed papers appear (Mind2Dialogue and the LLM scientific-law-discovery benchmark).
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 statistics come from an OpenAlex snapshot only (identity: openalex); citation counts, h-index, and per-year tallies are not independently verified.
- The reproducibility score averages just one scored paper (reproMean 20, reproPapers 1), so the 20/100 value is weakly grounded.
- Output and impact dimensions are unscored because the underlying data is insufficient; no estimates are provided.
- 'Letian Peng' may refer to multiple researchers; this profile is name-resolved from public bibliographic data, and the two Evolve-reviewed papers are linked via the OpenAlex identity.