Prolific machine-learning researcher bridging LLM interpretability and numerical decoding with spatiotemporal prediction
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
27
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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—
Not scored: this snapshot carries no per-venue publication counts or venue-quality data for the output axis. It would measure how much and where the researcher publishes, such as total works, venue tier, and first-author share — figures not available here.
Impact—
Not scored: although an h-index of 27 and i10-index of 57 are recorded, the citation distribution needed for the impact axis is unavailable in this snapshot. It would measure field influence via per-paper citations and downstream method adoption.
Focus0/100
Focus scores 0/100, the lowest concentration reading in this snapshot. OpenAlex assigns 70 works to traffic prediction and management, 59 to human mobility and location-based analysis, 40 to data management and algorithms, 23 to graph neural networks, and 22 to time series analysis, while the reviewed field distribution splits evenly (0.5/0.5) between LLM numerical readout and mechanistic interpretability — two partly distinct research lines rather than one narrow agenda.
Collaboration39/100
Collaboration scores 39/100. The recent three-year window lists 105 works but only 6 distinct coauthors, a small steady circle — Weiwei Ye, Hangchen Liu, Wenkang Wei, Yuan Fang, Hong Cheng, and Xingtong Yu recur most often. On the two reviewed 2026 papers the researcher appears as last author on one and middle author on the other, with neither flagged as corresponding.
Activity100/100
Activity scores 100/100, the ceiling of this radar. Annual works climb from 1 in 2014 to 5 in 2018, 20 in 2022, 30 in 2024, and 38 in 2025, with 37 already recorded for 2026 and 105 works across the recent three-year window — a steep, still-rising publication tempo concentrated in the last five years.
Reproducibility55/100
Reproducibility scores 55/100, a moderate mid-range reading computed from 2 repro-assessed papers with a mean of 55. Both reviewed papers — the NumericJev training-free decoding pipeline and the layerwise intervention study of internal knowledge retrieval — carry a medium repro-readiness rating, suggesting workable but non-trivial replication effort around hosted models and intervention setups.
Research summary
Renhe Jiang (Tokyo University of the Arts; ORCID 0000-0003-2593-4638) presents a bibliometric record of steeply accelerating output: h-index 27, i10-index 57, and 105 works in the recent three-year window, with annual counts rising from 1 (2014) to 30 (2024), 38 (2025), and 37 already logged for 2026. OpenAlex topic data places the bulk of this record in traffic prediction and management (70 works), human mobility and location-based analysis (59), data management and algorithms (40), graph neural networks (23), and time series forecasting (22). The two papers Evolve has reviewed sit elsewhere on this map: both are 2026 arXiv preprints in LLM research, where the profile's field distribution splits evenly between mechanistic interpretability ('From Parameters to Answers') and numerical readout (NumericJev, training-free multiway-decision-tree decoding). Whether this reflects a deliberate recent pivot toward LLM internals or an artifact of name disambiguation cannot be settled from the snapshot; the divergence is flagged in the caveats. Radar scores capture the shape of the record: activity 100/100, collaboration 39/100 on a small recurring co-author circle (6 distinct in three years), focus 0/100 reflecting wide topical spread, and reproducibility 55/100 from two medium-readiness papers, while output and impact remain unscored pending citation-level data.
Collaboration pattern
The collaboration footprint is narrow but stable: 6 distinct coauthors over the recent three-year window (105 works), matching a radar collaboration score of 39/100. Recurring partners include Weiwei Ye, Hangchen Liu, Wenkang Wei, Yuan Fang, Hong Cheng, and Xingtong Yu. On the two reviewed 2026 papers the researcher takes the last author position on NumericJev and a middle position on 'From Parameters to Answers', and is not flagged as corresponding author on either.
Research keywords
Highlights
Training-free numeric output from choice-only LLMs
Layerwise account of how LLMs retrieve internal knowledge
Output accelerating to 38 works in a single year
Annual publication counts rise from 1 (2014) to 20 (2022), 30 (2024), and 38 (2025), with 37 already recorded for 2026 and 105 works in the recent three-year window — the fastest-moving signal in this profile and the basis of its 100/100 activity score.
Timeline
- 2014First recorded work appears in the bibliometric record, a single publication.
- 2021Output reaches 9 works, continuing the steady climb through 2020's 7.
- 2022Annual output jumps to 20 works, doubling the prior year's count.
- 2025Most productive recorded year: 38 works, following 30 in 2024.
- 2026Two reviewed arXiv papers appear — NumericJev (last-author position) and 'From Parameters to Answers' (middle position) — marking the LLM-focused strand; 37 works are recorded for the year.
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.
- All bibliometrics derive from an OpenAlex-based snapshot (identity: openalex); no other databases were consulted.
- The output and impact radar axes are unscored because venue-level and citation-distribution data are absent; only h-index 27 and i10-index 57 are on record.
- Profile is name-resolved from public bibliographic data: the OpenAlex topic profile (traffic prediction, human mobility) diverges from the reviewed 2026 LLM papers, which may reflect a research pivot or author-disambiguation mixing.
- The reproducibility score rests on only 2 repro-assessed papers with medium readiness, so the 55/100 reading is thin evidence.