AI evaluation researcher: auditing LLM multi-agent social simulation and benchmarking vision-language models
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—
No publication-count, venue, or productivity data is available yet, so this axis is unscored. It would measure total research output — the number of papers, their publication venues, and the cadence of output over time.
Impact—
Citation counts, h-index, and downstream adoption are absent from the current snapshot, so impact is unscored. This axis would measure how often and how widely the work is cited, reused, or built upon by other researchers.
Focus0/100
The focus score is 0, reflecting an even 50/50 split in the field distribution between vision-language model / multimodal AI evaluation and AI evaluation / multi-agent systems (LLM social simulation). The two cataloged papers sit in these distinct clusters, so the profile currently reads as evaluation-driven and breadth-first rather than concentrated in a single research line.
Collaboration53/100
The collaboration score is 53, based on 13 distinct co-authors across the two cataloged 2026 papers. Recurring names include Haonan Jiang, Guojian Zhan, Jiancong Xie, Shijun Wan, and Yahui Liu, indicating team-based work rather than solo authorship; both papers place the researcher in a middle author position without corresponding-author status.
Activity—
No per-year publication timeline is available beyond the two 2026 preprints, so research activity is unscored. It would measure ongoing output pace — papers per year and continuity of publishing across recent years.
Reproducibility55/100
Reproducibility averages 55 across 2 scored papers (reproMean 55), with both rated medium repro-readiness. NoteVQA ships a concrete 252-item benchmark that supports replication, though self-judging via a Qwen backbone and a small sample temper it; the theory-of-mind audit is methodologically sound but carries a single-reader caveat.
Research summary
Cheng Chen is an early-career AI evaluation researcher whose cataloged work spans two complementary directions: auditing multi-agent LLM systems and benchmarking vision-language models. The Evolve catalog holds two 2026 arXiv preprints, both with middle-author, non-corresponding roles. 'Mind or Message? Auditing Theory of Mind in Multi-Agent Social Simulation' probes whether LLM agents in multi-issue negotiation track minds or merely messages, finding fluent but projection-driven, frontier-blind behavior, with egocentric projection examined against Pareto frontier outcomes. 'NoteVQA' builds a 252-item benchmark of real-life questions drawn from human communities, where the best-performing frontier VLM reaches only 52.8%, evaluated with agentic retrieval and interleaved multimodal answers under an LLM-as-judge protocol. The field distribution splits evenly (50/50) between vision-language/multimodal AI evaluation and AI evaluation/multi-agent systems (LLM social simulation), yielding a focus score of 0 and an evaluation-centric, breadth-first profile. Collaboration is team-based: 13 distinct co-authors across two papers and a collaboration score of 53. Reproducibility averages 55 over 2 scored papers, both rated medium repro-readiness, with concrete benchmarks offset by caveats such as self-judging via a Qwen backbone, small sample sizes, and single-reader audits. Output, impact, and activity metrics are not yet available. No affiliation or ORCID is on record; the profile is name-resolved from public bibliographic data.
Collaboration pattern
Work is collaborative: the two 2026 catalog papers together involve 13 distinct co-authors, with the researcher in middle-author, non-corresponding roles. Recurring collaborators in the network include Haonan Jiang, Guojian Zhan, Jiancong Xie, Shijun Wan, Dongiia Zhao, Yahui Liu, Chuan Mu, Cong Li, Thomas Fung, Alex Rossi, and Yi Li. Note that the co-author list itself contains a 'Cheng Chen' entry, likely a name-resolution artifact.
Research keywords
Highlights
NoteVQA: everyday questions that stump frontier VLMs
Auditing theory of mind in LLM multi-agent negotiation
Timeline
- 2026Published the 'Mind or Message?' arXiv preprint, an audit of theory of mind in multi-agent LLM social simulation, as a middle author.
- 2026Published the NoteVQA arXiv preprint, a 252-item benchmark of real-life visual questions from human communities, as a middle author.
- 2026Both cataloged papers appeared in 2026 across two distinct evaluation areas, marking an entry into AI evaluation research alongside 13 distinct co-authors.
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.
- Cheng Chen is a common name; this profile is name-resolved from public bibliographic data, and no ORCID, affiliation, or country was available to confirm identity.
- Bibliometric scores come from the platform's public bibliometric snapshot only; no citation counts, venue records, or h-index data were included, so output, impact, and activity remain unscored.
- The evidence base is two 2026 arXiv preprints with middle-author, non-corresponding roles; earlier or later publications are not represented in this snapshot.
- The frequent-coauthor list includes a 'Cheng Chen' entry alongside the researcher, likely a disambiguation artifact, so the distinct-coauthor count of 13 warrants caution.