NLP researcher at McGill University working on LLM agents, automated research, and contextual causality
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
Publication-volume data for this researcher is not yet available in the bibliometric snapshot, so output is unscored. This axis would measure total publication count, venue spread, and authorship roles accumulated over time.
Impact—
Citation-based impact data is not yet available, so this axis is unscored. It would measure citations and downstream uptake of the two cataloged preprints once citation coverage accumulates.
Focus0/100
The focus axis scores 0, the lowest value on this radar. The underlying field distribution splits evenly — 0.5 to natural language processing (LLM agents / automated research) and 0.5 to natural language processing — across only two adjacent NLP subfields. The 0 likely reflects the even two-way split rather than wide thematic sprawl, but with only two cataloged papers the signal should be interpreted cautiously.
Collaboration36/100
Collaboration scores 36. Five distinct coauthors are observed across the cataloged work — Yiheng Zhao, Mengzhuo Chen, Pengyi Liao, Yihan Huang, and Jun Yan — indicating a small, stable writing team rather than a broad network. On both 2026 preprints Hu holds a non-corresponding role (last author on the causality survey, middle position on the creativity-metrics paper), consistent with contribution within a group effort.
Activity—
Temporal publishing data is insufficient, so activity is unscored. This axis would measure publication cadence and recency across years; only two 2026 preprints are currently on record, too few to establish a trend.
Reproducibility38/100
Reproducibility averages 38 across two assessed papers. The ratings differ by output type: the contextual-causality survey is rated low readiness, as narrative syntheses without a systematic search are hard to replicate, while the creativity-metrics paper reaches medium readiness thanks to interpretable valueness/novelty metrics — though its headline correlation (n=6, p=0.063) weakens statistical confidence.
Research summary
Chengming Hu is a natural language processing researcher at McGill University whose Evolve-cataloged work centers on two questions: how large language models reason about cause and effect in context, and how to measure the creativity of LLM-driven research agents. The current evidence base is two 2026 arXiv preprints. The first, a survey on contextual causality with large language models, positions Hu as last author and organizes the literature through a semantic/intervention/counterfactual taxonomy; Evolve credits its clear roadmap while noting the narrative synthesis lacks a systematic search. The second, on measuring the creativity of frontier LLMs in automated research, introduces interpretable valueness and novelty metrics; Evolve judges the framing sensible but flags that the headline correlation rests on n=6 with p=0.063. Bibliometrically the profile is early-stage: field distribution splits evenly at 0.5/0.5 between NLP broadly and LLM agents/automated research; five distinct coauthors yield a collaboration score of 36; reproducibility averages 38 across two assessed papers, mirroring readiness ratings of low and medium; the focus axis records 0. Output, impact, and activity remain unscored for lack of data.
Collaboration pattern
The observable pattern is a compact, recurring team: five distinct coauthors — Yiheng Zhao, Mengzhuo Chen, Pengyi Liao, Yihan Huang, and Jun Yan — account for the collaboration network, yielding a collaboration score of 36. Hu appears in non-corresponding roles on both 2026 preprints: last author on the contextual-causality survey and a middle position on the creativity-metrics paper. With only two cataloged papers, the network's true breadth may be understated.
Research keywords
Highlights
Taxonomy for contextual causality with LLMs
Interpretable metrics for LLM research creativity
Repro-readiness: low versus medium
Evolve's assessments assign low repro-readiness to the causality survey and medium to the creativity-metrics paper, averaging 38 across the two — a split that reflects how methodological transparency differs between narrative surveys and metric-driven studies.
Timeline
- 2026Serves as last author on the arXiv survey "Contextual Causality with Large Language Models", proposing a semantic/intervention/counterfactual taxonomy.
- 2026Co-authors "Measuring the Creativity of Frontier LLMs in Automated Research" (arXiv, middle position), introducing interpretable valueness and novelty metrics.
- 2026Both preprints enter the Evolve review catalog, receiving repro-readiness ratings of low and medium respectively.
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.
- Dimension scores are computed from a public bibliometric snapshot; the output, impact, and activity axes are unscored because the underlying data is insufficient.
- The evidence base is narrow: only two 2026 arXiv preprints are reviewed by Evolve, so this is an early-stage profile.
- Neither cataloged paper lists Hu as corresponding author, and authorship history beyond these two papers is not captured.
- The profile is name-resolved from public bibliographic data linked to ORCID 0000-0002-1099-0736; "Chengming Hu" is a shared name and conflation with other researchers is possible.