University of Manchester researcher probing LLM time-series understanding and AI credibility signals
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
Output is unscored: no publication-count data is available in the radar. This axis would measure total scholarly production — paper counts, venue types, and yearly volume. Current coverage includes only two 2026 arXiv papers, too thin to score.
Impact—
Impact is unscored: no citation or influence metrics are available in the data. This axis would measure how widely the work is cited and built upon. Both catalog papers date to 2026, too recent for citation evidence.
Focus100/100
The radar scores focus at 100, but this rests on a single classified paper: Evolve assigns exactly one field, natural language processing / LLM evaluation for time-series reasoning. Within that thin base concentration is total; the second 2026 paper on AI-based credibility indicators has no assigned field yet, so apparent specialization may partly reflect incomplete classification rather than a deliberate single-track agenda.
Collaboration55/100
Collaboration scores 55, grounded in 15 distinct coauthors in the bibliometric record. The frequent list spans Jie Gong, Maowei Jiang, Yankai Chen, Guojun Xiong, Xue Liu, Min Peng, Qianqian Xie, Sophia Ananiadou, Zhuoran Lu, Weilong Wang, and Yangyang Yu. Both catalog papers place this researcher in a middle position without corresponding-author role, a contributor-level pattern inside larger teams; note that a duplicate of the subject's own name appears in the list, suggesting possible namesake noise.
Activity—
Activity is unscored: no yearly publication-cadence data exists in the record. This axis would measure ongoing research rhythm — papers per year and sustained output over time. The visible record spans only 2026, so no trend can be established.
Reproducibility55/100
Reproducibility scores 55, computed as a mean over exactly one scored paper: TimeLitmus, whose Evolve readiness is rated medium. The benchmark's rule-based, counterfactual design aids reuse, but the verdict flags rule-generated labels and a small pair count as limitations. With a single observation, this score describes one artifact, not a body of reproducible work.
Research summary
Zhiwei Liu is a researcher at the University of Manchester whose Evolve record, name-resolved from public bibliographic data, sits at an unusual intersection: large language models as objects of time-series evaluation, and AI signals inside human information ecosystems. The catalog's reviewed entry, TimeLitmus (arXiv, 2026), is a diagnostic benchmark for cross-modal understanding and explanation faithfulness in event-conditioned time-series prediction; Evolve's verdict calls it a solid diagnostic benchmark finding that LLMs' event-conditioned accuracy overstates cross-modal skill, with the caveat of rule-based labels and a small number of pairs. A second 2026 arXiv paper, One AI Signal, Many Human Judgments, applies Bayesian cascade analysis to AI-based credibility indicators in online information spread, extending the same question — how machine outputs are interpreted — to social settings; it awaits Evolve review. Radar scores give focus 100 on a single classified field (natural language processing / LLM evaluation for time-series reasoning), collaboration 55 across 15 distinct coauthors, and reproducibility 55 averaged over one scored paper. Output, impact, and activity are unscored for lack of data. Both papers list a middle author position without corresponding role, so this reads as an early-stage, contributor-level profile.
Collaboration pattern
The bibliometric record shows 15 distinct coauthors, and the frequent list includes Jie Gong, Maowei Jiang, Yankai Chen, Guojun Xiong, Xue Liu, Min Peng, Qianqian Xie, Sophia Ananiadou, Zhuoran Lu, Weilong Wang, and Yangyang Yu. In both catalog papers this researcher sits in a middle author position and is never corresponding, consistent with contributing inside larger teams. A duplicate of the subject's own name in the coauthor list suggests namesake noise in the underlying data.
Research keywords
Highlights
TimeLitmus: LLM accuracy overstates cross-modal skill
Counterfactual probes for explanation faithfulness
One AI signal, many human judgments
A second 2026 arXiv paper applies Bayesian cascade analysis to AI-based credibility indicators in online information spread, asking how a single machine signal is absorbed across human judgments; it is not yet Evolve-reviewed, so no verdict or keywords exist.
Timeline
- 2026TimeLitmus appears on arXiv: a diagnostic benchmark for cross-modal understanding and explanation faithfulness in event-conditioned time-series prediction.
- 2026One AI Signal, Many Human Judgments appears on arXiv: a Bayesian cascade analysis of AI-based credibility indicators in online information spread.
- 2026Both catalog papers fall in 2026, marking the visible start of this researcher's Evolve record; output, impact, and activity metrics are not yet available.
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.
- Radar scores are computed from public bibliometric data only; output, impact, and activity lack sufficient data and are left unscored rather than estimated.
- "Zhiwei Liu" is a common name; this profile is name-resolved from public bibliographic data, and a duplicate of the subject's own name appears in the frequent-coauthor list, so the coauthor network may mix namesakes.
- Only TimeLitmus carries an Evolve verdict, keywords, and repro-readiness rating; the Bayesian cascade paper has no field, keywords, verdict, or slug assigned yet.
- No citation counts, h-index, venue history, or career facts are included in this snapshot; the record covers only two 2026 arXiv papers.