Harbin Institute of Technology researcher working on LLM agents, VLM interpretability, and embodied AI evaluation.
3
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
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—
Not yet scored: the snapshot lacks complete production statistics for this researcher. This axis would measure cumulative work counts, year-by-year output trajectory, and venue distribution — pending fuller bibliographic coverage beyond the 26 recent works already recorded.
Impact—
Not yet scored: citation counts are absent from this OpenAlex snapshot, where h-index and i10-index appear only as zero placeholders. This axis would measure citation influence, highly cited works, and field-normalized impact once citation data becomes available.
Focus0/100
Scores 0/100, which signals maximal topic breadth rather than a deficit: the field distribution splits evenly at 0.5/0.5 between LLM-based agents and multimodal vision-language interpretability, while the topic profile spans five clusters led by Topic Modeling (14), NLP techniques (10), multimodal ML applications (10), neural network applications (5), and RL in robotics (5).
Collaboration54/100
Scores 54/100, a moderate collaborative intensity: 14 distinct co-authors appear in the recent window, with recurring partners including Zeming Liu, Yutong Li, Wei Su, Heyan Huang, and Haifeng Wang. Across the three catalog papers this researcher holds middle authorship twice and last (senior) position once, without corresponding-author flags.
Activity96/100
Scores 96/100, the standout axis in the radar: 26 works fall in the recent three-year window, with 25 dated 2026 against just 1 in 2025 — a near-vertical ramp in output concentrated in the current year, spanning the agents, interpretability, and embodied evaluation lines.
Reproducibility55/100
Scores 55/100, computed from 2 papers assessed with a mean readiness of 55. All three catalog papers carry a medium reproducibility rating, and the review verdicts flag single-run evaluations, absent significance tests, and narrow model coverage as the main gaps to address.
Research summary
Wanxiang Che is a researcher at Harbin Institute of Technology, China, with a record name-resolved from OpenAlex bibliographic data. Recent work splits evenly — 0.5/0.5 in the field distribution — between natural language processing with LLM-based agents and machine-learning interpretability for multimodal vision-language models. The snapshot records 26 works in the recent three-year window, 25 of them dated 2026 against just 1 in 2025, producing a standout activity score of 96/100. Three 2026 arXiv papers reviewed by Evolve illustrate the span: a mechanistic activation-patching study of how vision-language models read values from vertical bar charts (last authorship), LifeMem, an agent-memory framework that clusters trajectories into reusable workflows, and a manual-free benchmark for visually grounded exploration of household devices. Review verdicts converge on a pattern: solid methodological craft — rigorous patching, a ten-environment study, rich diagnostics — offset by narrow model coverage, single-run evaluations, and missing significance testing. Reproducibility is scored 55/100 from two assessed papers, all rated medium readiness. Collaboration is moderate at 54/100 across 14 distinct recent co-authors; output and impact remain unscored because citation counts are absent from the snapshot.
Collaboration pattern
Collaboration is moderate (54/100): 14 distinct co-authors appear in the recent three-year window, with recurring partners including Zeming Liu, Yuli Qiu, Yutong Li, Wei Su, Heyan Huang, Haifeng Wang, and Yuhang Guo, spanning the agents and vision-language groups. On the three reviewed papers this researcher holds middle authorship twice and last (senior) position once, with no corresponding-author flags — a pattern consistent with co-supervision of parallel agent- and VLM-focused lines rather than solo work.
Research keywords
Highlights
How VLMs Read Bar Charts — a Patching Study
LifeMem: Lifelong Experience Reuse for LLM Agents
A Manual-Free Benchmark for Grounded Device Exploration
Timeline
- 2025First recorded output in the snapshot window appears: 1 work, preceding the 2026 surge.
- 2026Three Evolve-reviewed arXiv papers land: a VLM chart-reading interpretability study, the LifeMem agent-memory framework, and a grounded household-device exploration benchmark.
- 2026Recorded output peaks at 25 works in the bibliometric snapshot, driving the 96/100 activity score.
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 figures come from an OpenAlex snapshot only; no ORCID was available, so this profile is name-resolved from public bibliographic data.
- Citation metrics are missing: h-index and i10-index appear as zero placeholders, so output and impact are unscored rather than genuinely zero.
- Counts by year are available only for 2025 (1 work) and 2026 (25 works); earlier career output is not covered in this snapshot.
- Reproducibility scores and review verdicts cover only the 3 Evolve-reviewed 2026 arXiv preprints; venue acceptance and peer-review status are not recorded.