Qifan Wang: audio-language models for telecom fraud detection — TeleAntiFraud benchmark and FRAUDSkill (arXiv 2026)
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
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—
Not yet scored: this snapshot contains no publication-volume or productivity metrics beyond the two reviewed 2026 arXiv papers. This axis would measure cumulative output volume, venue distribution, and yearly production once a fuller bibliographic record is linked.
Impact—
Not yet scored: no citation counts, h-index, or influence indicators appear in the input data. This axis would measure citation uptake and downstream use of the TeleAntiFraud benchmark once citation records become available.
Focus0/100
The radar assigns focus a score of 0, reflecting a field distribution split exactly 50/50 between two labels: 'speech processing and NLP (audio-language model benchmarking for fraud detection)' and 'speech and audio machine learning (audio-language models for fraud detection)'. In substance these describe one tightly concentrated topic — audio-language models for fraud detection — so the zero likely reflects near-duplicate label fragmentation rather than genuine topical breadth; both reviewed papers sit squarely in this single area.
Collaboration66/100
Collaboration scores 58, grounded in 17 distinct coauthors in the snapshot, twelve of whom recur as frequent collaborators — including Huiyuan Liu, Shun Zhang, Yifan Wang, and Yijin Zhou. Both catalog papers carry middle authorship with no corresponding role, consistent with contributing inside a larger, stable team rather than solo or senior-led work; the moderate score reflects a real but not yet expansive collaborative footprint.
Activity40/100
Not yet scored: beyond two papers both dated 2026, there is no multi-year cadence or submission record in the input. This axis would measure recent production velocity and research tempo across years once a longer publication history is linked.
Reproducibility85/100
Reproducibility scores 78, the mean across two scored papers (reproPapers: 2). The TeleAntiFraud 2.0 benchmark is rated high readiness, consistent with its refreshable, profile-grounded design and collapse-aware analysis; FRAUDSkill is rated medium, and Evolve's ablation reading — most of the +31.96 Macro-F1 gain traced to structured post-processing rather than skill search — signals a method-level reproducibility risk to check before reuse.
Research summary
Qifan Wang is an emerging researcher at the intersection of speech and audio machine learning and NLP, focused on audio-language models for telecom fraud detection. This is an early-stage profile: the Evolve catalog holds two 2026 arXiv papers, both with middle authorship and no corresponding role, within a pool of 17 distinct coauthors, 12 of them frequent. The first, TeleAntiFraud 2.0, is a refreshable, profile-grounded, audio-based benchmark for telecom fraud detection; Evolve's review credits its near-domain negative design and collapse-aware analysis of shortcut learning and prediction collapse, while flagging synthetic-only data and thin human validation. The second, FRAUDSkill, proposes structured frozen-weight skill optimization for audio anti-fraud detection, reporting a +31.96 Macro-F1 gain on the TeleAntiFraud benchmark; Evolve's ablation reading attributes most of that gain to structured post-processing rather than the skill search itself — a caveat prospective users should weigh. Radar scores set reproducibility at 78 (the mean across two scored papers; the benchmark rates high readiness, the method paper medium) and collaboration at 58, while output, impact, and activity lack sufficient data and are left unscored. The profile is name-resolved from public bibliographic data; no institutional affiliation is available in this snapshot.
Collaboration pattern
The snapshot records 17 distinct coauthors, 12 of them recurring — Huiyuan Liu, Shun Zhang, Yifan Wang, Yijin Zhou, Yuxi Zhao, Chengxian Hu, Peng Chen, Zhiming Ma, Yanxing Liu, Di Liu, Yuyang Deng, and Haoyang Meng. Both papers carry middle authorship without a corresponding role, suggesting a contributor within a larger, stable fraud-detection team whose papers pair benchmark construction with method evaluation, rather than a solo or corresponding-author lead.
Research keywords
Highlights
Refreshable audio benchmark for telecom fraud detection
Frozen-weight skill optimization for audio anti-fraud
Benchmark-plus-method loop on a single task
The two reviewed papers pair a benchmark (TeleAntiFraud 2.0) with a method (FRAUDSkill) evaluated on that same benchmark, inside one 17-collaborator network — a compact, self-referential evaluation loop concentrated on audio-language models for fraud detection.
Timeline
- 2026Released TeleAntiFraud 2.0 on arXiv: a refreshable, profile-grounded, audio-based benchmark for telecom fraud detection with near-domain negatives and collapse-aware analysis (Evolve reproducibility readiness: high).
- 2026Released FRAUDSkill on arXiv: structured frozen-weight skill optimization for audio anti-fraud detection, reporting +31.96 Macro-F1 on TeleAntiFraud, with most of the gain traced to structured post-processing (reproducibility readiness: medium).
- 2026Both releases sit in a 17-distinct-collaborator network (12 frequent coauthors) with middle, non-corresponding authorship, anchoring radar reproducibility at 78 and collaboration at 58.
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.
- 'Qifan Wang' is a common name; this profile is name-resolved from public bibliographic data only — no ORCID or institutional affiliation was provided, so identity confidence is limited.
- Radar scores and the 17-coauthor count come from an OpenAlex-derived bibliometric snapshot; only two reviewed papers (both 2026 arXiv preprints) are in scope.
- Output, impact, and activity dimensions lack underlying data and are left unscored; no citation counts or publication venues are available.
- The focus score of 0 stems from a 50/50 split between two near-duplicate field labels, not from verified topical breadth.