Researcher at the Chinese Academy of Sciences working across adversarial ML and few-shot audio learning.
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-14) 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: this snapshot records h-index and i10-index as zero with no reliable cumulative publication totals, so the output axis — measuring the volume and consistency of research output over time — has insufficient data to compute.
Impact—
Not yet scored: citation counts are unavailable and h-index stands at 0 in the snapshot, so the impact axis — measuring citation-based influence and uptake of the work by other researchers — has no basis for a score.
Focus0/100
The focus axis scores 0, the lowest in this profile, and the raw data supports reading it as genuinely broad rather than missing: OpenAlex splits the field distribution exactly evenly at 0.5 between computer-vision adversarial machine learning and audio signal processing, and topic counts spread across five areas — Domain Adaptation and Few-Shot Learning (7), Multimodal ML Applications (5), Neural Network Applications (4), Image Enhancement (3), and Image/Video Retrieval (2).
Collaboration59/100
The collaboration axis scores 59. The OpenAlex snapshot records 18 distinct coauthors, with Qi Ming, Juan Fang, and Jiahuan Zhou appearing repeatedly in the frequent-coauthor list, indicating a small set of recurring partners around a wider network. Both cataloged papers are multi-author efforts, with middle authorship on the visible-infrared attack work and last authorship on the audio few-shot paper, but neither lists this researcher as corresponding author.
Activity75/100
The activity axis scores 75, backed by strong recent volume: countsByYear records 2 works in 2025 rising to 10 in 2026, and the snapshot lists 12 works over the last three years. All recorded output sits in the two most recent years, so the score reflects a steeply ramping publication cadence on a short record rather than sustained long-run volume.
Reproducibility78/100
The reproducibility axis scores 78, computed from Evolve's review of 2 papers (reproMean 78). The audio few-shot paper is rated high reproducibility readiness, with a 4.9-8.4 point gain over a matched control, though on a single benchmark without confidence intervals; the visible-infrared attack paper is rated medium, with broad results but no physical-world validation.
Research summary
Jiangmeng Li is a researcher at the Chinese Academy of Sciences whose cataloged work spans two distinct areas of machine learning: adversarial machine learning in computer vision and audio signal processing. This profile is name-resolved from public OpenAlex bibliographic data, which splits the field distribution exactly evenly (0.5/0.5) across those two areas and lists five topic clusters led by Domain Adaptation and Few-Shot Learning (7 papers), Multimodal Machine Learning Applications (5), and Advanced Neural Network Applications (4). Two Evolve-reviewed arXiv papers, both from 2026, anchor the qualitative record. The first, 'Stealthy in Semantics, Antagonistic in Space', introduces CamoShift, an object-level misalignment attack that camouflages a patch and digitally shifts infrared evidence against visible-infrared object detectors; Evolve judged the results strong and broad but flagged the absence of physical-world validation. The second, 'Sample-Conditioned Representation Selection for Audio Few-Shot Learning', applies Gumbel Top-k per-input channel selection and improves OOD few-shot audio accuracy by 4.9-8.4 points over a matched control, on a single benchmark without confidence intervals. Activity is the strongest measured dimension (75), with 12 works in the last three years and 10 recorded in 2026. Citation-based output and impact are not yet scored because the snapshot records h-index and i10-index as zero.
Collaboration pattern
The collaboration axis scores 59. The snapshot records 18 distinct coauthors, with Qi Ming, Juan Fang, and Jiahuan Zhou each appearing twice in the frequent-coauthor list — the clearest signal of recurring partners. Both 2026 catalog papers are multi-author efforts in which Jiangmeng Li is neither first nor corresponding author (middle position on the visible-infrared attack paper, last position on the audio few-shot paper), suggesting a contributor role within team-based projects rather than solo leadership.
Research keywords
Highlights
CamoShift: object-level misalignment attacks on visible-infrared detectors
Gumbel Top-k channel selection lifts OOD few-shot audio accuracy
Early-stage record with an above-average reproducibility signal
With only 2 Evolve-reviewed papers in a record starting in 2025, this profile is early-stage, yet both papers carry actionable reproducibility signals (one high, one medium) yielding a reproducibility mean of 78 — tied with activity (75) as the strongest scored dimensions.
Timeline
- 2025Earliest works in this OpenAlex snapshot: 2 publications recorded, opening a short but fast-rising record.
- 2026Output accelerates to 10 recorded works; both Evolve-reviewed arXiv papers fall in this year.
- 2026Co-authored the visible-infrared object detector attack paper 'Stealthy in Semantics, Antagonistic in Space' (middle authorship).
- 2026Last author on 'Sample-Conditioned Representation Selection for Audio Few-Shot Learning'.
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 statistics come from an OpenAlex snapshot only; h-index and i10-index are recorded as 0, so the output and impact axes are unscored rather than low.
- Qualitative verdicts rest on just two Evolve-reviewed papers, both 2026 arXiv preprints; no peer-reviewed venue information is included.
- No ORCID is provided; the profile is name-resolved from public bibliographic data and may conflate namesakes.
- The reproducibility mean (78) is computed from only 2 reviewed papers and may not generalize.