Jiayao Liu

Beijing Institute of Technology · China

Unclaimed · AI-structured profileOpenAlex

Jiayao Liu (Beijing Institute of Technology): zero-shot anomaly detection and neural video compression.

papers here

reviewed by Evolve

first-author

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 scored: the underlying output data is insufficient in this snapshot. This axis would measure total publication volume and its growth across the career; only partial per-year counts and two reviewed preprints are on record.

Impact—

Not scored: citation-based impact data is unavailable here. This axis would measure how widely the work is cited and taken up by other researchers; available signals (h-index 4, i10-index 1) are too early-stage to summarize.

Focus0/100

Score 0, indicating a deliberately split profile rather than a single lane. The field distribution divides evenly at 0.5 between computer vision (anomaly detection) and computer vision (neural video compression), and topic clusters span five areas — advanced neural network applications (5), anomaly detection techniques (4), human pose and action recognition (3), smart agriculture and AI (2), and infrastructure maintenance and monitoring (2).

Collaboration41/100

Score 41. The OpenAlex record lists 7 distinct co-authors in the recent window — Xuezhi Xiang, Heqi Xiang, Shanjun Zhang, Guanghao Wu, Xiaoheng Li, Yiming Chen, and Yixin Zhao — pointing to a compact, recurring team rather than a wide network. Both reviewed 2026 preprints place the researcher in a middle author position, consistent with a steady contributing role inside that group.

Activity79/100

Score 79, the strongest axis here. OpenAlex records 14 works in the most recent three-year window, with yearly counts rising from 2 in 2022 to 5 in 2023, 6 in 2024, 3 in 2025, and 5 already in 2026 — a sustained, generally upward cadence of new computer-vision output.

Reproducibility55/100

Score 55, averaged over 2 reviewed papers, both rated medium reproducibility readiness. Evolve notes that PSMP-CLIP delivers ablated pixel-level gains but leaves key hyperparameters unreported and ships no code, while TCNeRV is judged credible yet validated on a single dataset (UVG) with gains fading at the 10M-parameter scale.

Research summary

This profile is name-resolved from OpenAlex bibliographic records for Jiayao Liu of Beijing Institute of Technology, China; the name is common, so identity should be verified independently. The record describes an early-stage computer-vision researcher whose work divides evenly between two subfields — anomaly detection (0.5) and neural video compression (0.5) — with topic clusters spanning advanced neural network applications (5), anomaly detection techniques (4), human pose and action recognition (3), smart agriculture and AI (2), and infrastructure maintenance and monitoring (2). Activity is the strongest axis: 14 works in the most recent three-year window, with yearly counts rising from 2 in 2022 to 6 in 2024 and 5 already recorded for 2026. Citation signals remain small (h-index 4, i10-index 1), so impact is not scored. Evolve has reviewed two 2026 arXiv preprints with the researcher in a middle author position: PSMP-CLIP, a CLIP+SAM2 zero-shot anomaly detector with ablated pixel-level gains on 7 of 11 benchmarks but no released code, and TCNeRV, a dual-domain temporal-context implicit neural video codec reporting +2.2 dB over HNeRV-Boost on UVG at 3M parameters. Both are rated medium reproducibility (mean 55 across 2 papers), and the collaboration footprint is compact: 7 distinct recent co-authors.

Collaboration pattern

The OpenAlex record shows 7 distinct co-authors in the recent window, and the frequent-coauthor list — Xuezhi Xiang, Heqi Xiang, Shanjun Zhang, Guanghao Wu, Xiaoheng Li, Yiming Chen, and Yixin Zhao — points to a compact, recurring team. The collaboration score of 41 reflects a small stable group rather than a broad network; on both reviewed 2026 preprints the researcher sits in a middle author position, consistent with a contributing member of an established team.

Research keywords

zero-shot anomaly detectionvision-language modelsCLIPSAM2prompt learninganomaly segmentationimplicit neural representationneural video codecvideo compressiontemporal context modelingfeature fusionresidual coding

Highlights

PSMP-CLIP: patch-prompt SAM and multi-semantic prompting for zero-shot anomaly detection
Evolve's review describes a CLIP+SAM2 zero-shot anomaly detector whose ablations show pixel-level gains on 7 of 11 benchmarks; image-level performance still lags rivals, key hyperparameters are unreported, and no code is released, hence the medium reproducibility rating.
TCNeRV: dual-domain temporal context modeling for implicit neural video compression
The reviewed preprint adds dual-domain temporal context to implicit neural representation video coding, reporting +2.2 dB over HNeRV-Boost on UVG at 3M parameters; Evolve judges the work credible, though it is validated on a single dataset and gains fade at the 10M-parameter scale.
Consistent middle-author contributor across both reviewed 2026 preprints
On both reviewed 2026 arXiv papers the researcher appears in a middle author position rather than first or corresponding author, consistent with the compact cluster of 7 distinct recent co-authors recorded in the OpenAlex profile.

Timeline

2022Earliest year in the bibliometric snapshot: 2 works recorded in the OpenAlex counts-by-year series.
2023Output rises to 5 works as the researcher's publication cadence accelerates.
2024Peak year within the provided window: 6 works recorded.
20253 works recorded in the yearly counts.
20265 works recorded so far, including two arXiv preprints reviewed by Evolve — PSMP-CLIP (zero-shot anomaly detection) and TCNeRV (implicit neural video compression) — both in middle-author positions.

Papers in the Evolve catalog

Frequent co-authors

Xuezhi Xiang×2 Heqi Xiang×2 Shanjun Zhang×2 Guanghao Wu×1 Xiaoheng Li×1 Yiming Chen×1 Yixin Zhao×1

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 metrics are computed from OpenAlex records only (identity: openalex); no ORCID is linked and the name is common, so the profile is name-resolved and should be verified independently.
Output and impact dimensions are unscored because the underlying data is insufficient in this snapshot; beyond h-index 4 and i10-index 1, no citation totals are available.
Only the 2 reviewed 2026 arXiv preprints carry qualitative Evolve verdicts; other works counted in the bibliography are not individually reviewed.
The researcher appears in middle-author positions on both reviewed papers; specific role and seniority cannot be inferred from this data.