Hongyu Wang

Beijing University of Posts and Telecommunications

Unclaimed · AI-structured profileOpenAlex

BUPT researcher at the LEO satellite crossroads: LLM-driven QoS routing and task-oriented semantic transmission.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-02) 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—

No output score yet: publication counts and venue data in the current snapshot are insufficient. This axis would measure cumulative and per-year paper volume across the researcher's full publication record.

Impact—

No impact score yet: citation counts and influence metrics are unavailable in the current data. This axis would measure citations to the work, their accumulation over time, and downstream scholarly influence.

Focus0/100

The focus score is 0, grounded in a field distribution split exactly 50/50 between computer networks (LEO satellite communications) and satellite remote sensing / semantic communications at the computer-vision-plus-wireless intersection. The reviewed output runs as two parallel threads rather than one narrow specialty, though both are anchored in satellite systems.

Collaboration39/100

Collaboration scores 39, computed from 6 distinct co-authors in the available data. Mugen Peng, Wenjia Xu, Bowen Lu, Yaohua Sun, Kerui Guo, and Shuoyuan Sun recur across the reviewed work, indicating a stable single-team pattern centered on BUPT with no cross-institution breadth visible in this snapshot. Both papers place Wang as a middle, non-corresponding author, consistent with contributing within an established group.

Activity—

No activity score yet: recent publication cadence cannot be computed from the available snapshot. This axis would measure ongoing output and research momentum in the most recent years of the record.

Reproducibility55/100

Reproducibility scores 55, the mean over the 2 papers Evolve assessed, both rated medium reproducibility readiness. STR-Agent reports large routing gains (45% to 92% intent fulfillment, roughly 60% delay cut) but rests on synthetic evaluation with weak baselines; the semantic transmission paper shows convincing low-SNR gains though its baselines lack task-aware training. Replication is plausible but needs careful reconstruction of the evaluation settings.

Research summary

Hongyu Wang is a researcher at Beijing University of Posts and Telecommunications whose reviewed output sits at the intersection of LEO satellite networking and semantic communications. Two 2026 arXiv papers define the current profile. The first, STR-Agent, is an LLM-driven agent that turns natural-language requests into QoS-aware routing in LEO satellite networks; Evolve's review records large gains — intent fulfillment rising from 45% to 92% and roughly a 60% delay reduction — while flagging synthetic evaluation and weak baselines. The second transmits frozen-backbone features through a tiny convolutional channel adaptation module for multi-task satellite remote sensing over low-SNR channels, showing large low-SNR gains over reconstruction-based joint source-channel coding, with baselines that lack task-aware training. Both papers list Wang as a middle author without corresponding role, consistent with a contributing position inside a larger BUPT team whose recurring co-authors include Mugen Peng, Wenjia Xu, Bowen Lu, Yaohua Sun, Kerui Guo, and Shuoyuan Sun. The recorded field distribution splits evenly between LEO satellite communications and satellite remote sensing / semantic communications. Output, impact, and activity scores are not yet computable from the available bibliometric data, so the profile should be read as an early snapshot of an active contributor rather than a complete record.

Collaboration pattern

The snapshot shows a compact, single-team pattern: 6 distinct co-authors, with Mugen Peng, Wenjia Xu, Bowen Lu, Yaohua Sun, Kerui Guo, and Shuoyuan Sun recurring across both reviewed 2026 papers. Wang appears as a middle author without corresponding role on either, consistent with contributing within a BUPT-centered group rather than leading an independent collaboration network; cross-institution partnerships are not visible in the available data.

Research keywords

LLM agentLEO satellite networksQoS-aware routingintent-based networkingsupervised fine-tuningtask-oriented semantic communicationjoint source-channel coding (JSCC)satellite remote sensinglow-SNR robustnessmulti-task pretrainingchannel adaptation module

Highlights

STR-Agent: LLM agent turns natural language into QoS-aware LEO routing
In a 2026 arXiv paper, Wang co-authored STR-Agent, an LLM-driven agent that converts natural-language requests into QoS-adaptive routing for LEO satellite networks using supervised fine-tuning over a Dijkstra routing base. Evolve's review notes large gains — intent fulfillment from 45% to 92% and roughly a 60% delay cut — capped by synthetic evaluation and weak baselines; reproducibility readiness is medium.
Task-oriented semantic feature transmission over low-SNR satellite channels
The second 2026 arXiv paper transmits frozen-backbone features through a tiny convolutional channel adaptation module for multi-task satellite remote sensing over low-SNR channels, showing large low-SNR gains over reconstruction-based JSCC. Evolve finds the results convincing while noting the baselines lack task-aware training; reproducibility readiness is medium.
One direction, two layers: learned intelligence across the satellite stack
Together, the two 2026 papers sketch a single agenda — bringing learned, intent-aware intelligence into LEO satellite systems — with STR-Agent acting at the network/routing layer and the semantic transmission work at the sensing/transmission layer, mirroring the recorded 50/50 split in field distribution.

Timeline

2026Co-authored (middle position) STR-Agent, an LLM-driven agent for QoS-aware routing in LEO satellite networks, released on arXiv.
2026Co-authored (middle position) the task-oriented semantic feature transmission paper for multi-task satellite remote sensing over low-SNR channels, released on arXiv.
2026Both papers were reviewed by Evolve with medium reproducibility readiness, establishing a two-thread profile across LEO networking and semantic communications.

Papers in the Evolve catalog

Frequent co-authors

Mugen Peng×2 Wenjia Xu×2 Bowen Lu×1 Yaohua Sun×1 Kerui Guo×1 Shuoyuan Sun×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.

Dimension scores come from a limited public bibliometric snapshot; output, impact, and activity lack sufficient underlying data and are left unscored.
Coverage is limited to two 2026 arXiv preprints reviewed by Evolve; peer-reviewed venue publications, if any, are outside this snapshot.
'Hongyu Wang' is a common name; this profile is name-resolved from public bibliographic data tied to the stated affiliation, and the Chinese character form of the name is unconfirmed.
Both records list middle, non-corresponding authorship, so individual contribution within the team cannot be precisely attributed.