Audio-language model researcher building benchmarks and adaptation methods for telecom fraud detection
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-19) 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: Evolve lacks reliable publication-volume data for this record (total outputs, venue types, first/corresponding-author counts). This axis would measure research productivity over time; only two reviewed 2026 preprints are on record, so no estimate is made.
Impact—
Not yet scored: no citation counts, h-index, or usage signals are available for this name. This axis would measure the influence of the work on subsequent research; the two preprints are too recent for meaningful citation evidence.
Focus0/100
Evolve's focus score is currently 0, which should be read against sparse data rather than as scattered interests. The raw field distribution splits evenly (0.5 and 0.5) between two labels that both describe audio-language models for fraud detection — one from speech processing and NLP, one from speech and audio machine learning. Both reviewed 2026 papers sit in this single problem area, so topical concentration is evident even though the scored value is 0.
Collaboration58/100
The collaboration score is 58. The record lists 17 distinct coauthors, with 12 recurring names — including Huiyuan Liu, Shun Zhang, Qifan Wang, and Yijin Zhou — who recur across the two 2026 fraud-detection papers. On both catalog papers Yifan Wang holds a middle authorship position and is not the corresponding author, suggesting a contributor role within a larger group rather than an independent collaboration hub.
Activity—
Not yet scored: no year-by-year publication volume or recency data underpins this axis. It would measure ongoing research tempo over time. The record shows only two 2026 preprints, so the current activity level cannot be responsibly estimated.
Reproducibility78/100
The reproducibility score is 78, averaged over two papers. TeleAntiFraud 2.0 is rated high for reproduction readiness — a refreshable, profile-grounded audio benchmark with near-domain negatives — while FRAUDSkill is rated medium. Evolve's review of FRAUDSkill cautions that most of the reported +31.96 Macro-F1 gain comes from structured post-processing, a nuance that matters for anyone reproducing the pipeline.
Research summary
Yifan Wang is an early-stage researcher in speech and audio machine learning whose reviewed work centers on audio-language models for telecom fraud detection. The Evolve catalog contains two 2026 arXiv preprints. The first, TeleAntiFraud 2.0, is a refreshable, profile-grounded, audio-based benchmark for telecom fraud detection; Evolve's review credits its near-domain negatives and collapse-aware analysis while noting that its data is synthetic-only and human validation is thin. The second, FRAUDSkill, applies structured frozen-weight skill optimization to audio anti-fraud detection on the TeleAntiFraud benchmark; the review attributes most of its reported +31.96 Macro-F1 gain to structured post-processing rather than the skill search itself. Bibliometric signals are limited but coherent: the field distribution splits evenly (0.5/0.5) between two labels that both describe audio-language model work for fraud detection, 17 distinct coauthors appear across the record with a collaboration score of 58, and a reproducibility score of 78 is averaged over two papers, one rated high and one medium for reproduction readiness. On both catalog papers Wang is a middle-position author rather than first or corresponding, so leadership and independence signals remain unestablished. Output, impact, and activity dimensions are currently unscored due to insufficient data, and no citation or productivity metrics are on record.
Collaboration pattern
With 17 distinct coauthors and 12 recurring names, Yifan Wang works within a compact, stable team rather than a wide open network — reflected in a collaboration score of 58. Recurring collaborators include Huiyuan Liu, Shun Zhang, Qifan Wang, Yijin Zhou, Yuxi Zhao, and Chengxian Hu. On both 2026 catalog papers Wang is a middle-position author and not corresponding, consistent with a specialist contributor role inside a fraud-detection research group.
Research keywords
Highlights
TeleAntiFraud 2.0: a refreshable, audio-based fraud benchmark
FRAUDSkill: frozen-weight skill optimization, with a caveat
A paired benchmark-plus-method research line
Timeline
- 2026Published TeleAntiFraud 2.0, a refreshable, profile-grounded, audio-based benchmark for telecom fraud detection (arXiv preprint, middle author).
- 2026Published FRAUDSkill, structured frozen-weight skill optimization for audio anti-fraud detection, evaluated on the TeleAntiFraud benchmark (arXiv preprint, middle author).
- 2026Recorded 17 distinct coauthors across the two fraud-detection papers, with a collaboration score of 58 and a reproducibility score of 78 averaged over the two papers.
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.
- Yifan Wang is a common name; this profile is name-resolved from public bibliographic data and may conflate multiple researchers.
- Bibliometrics (coauthor counts, field distribution, reproducibility scores) come from Evolve's OpenAlex-based pipeline; institution, country, and ORCID are missing from the record.
- Only two arXiv preprints (both 2026) are reviewed; peer-review status, citations, output volume, impact, and activity metrics are unavailable, so output, impact, and activity are unscored.
- Evolve's reviews note that TeleAntiFraud 2.0 uses synthetic-only data with thin human validation, and that FRAUDSkill's headline gain is largely post-processing — interpret results with these qualifications.