Benchmark builder at Alibaba for multilingual LLM-agent and long-form audio-language model evaluation
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-09) 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—
Insufficient data: this axis measures total publication volume and its year-by-year trajectory, but the snapshot provides only partial signals (13 works in the recent three-year window, h-index 1) without a complete indexed record, so no score is assigned yet.
Impact—
Insufficient data: this axis measures citations and downstream uptake of the work. The snapshot records h-index 1 and i10-index 0, which are too sparse to score meaningfully, so impact is left unscored rather than estimated from thin evidence.
Focus0/100
Scores 0/100: the OpenAlex field distribution splits evenly at 0.5 and 0.5 between audio-language model evaluation and LLM agent evaluation, and five topic clusters (Topic Modeling 14, NLP techniques 11, few-shot 5, multimodal 4, syntax and semantics 1) span both sides, so topical concentration is currently minimal.
Collaboration56/100
Scores 56/100: the snapshot lists 16 distinct coauthors and 12 recurring partners, including Zeyu Yang, Xinyu Zhang, Pei Zhang, and Satoshi Nakamura, spanning NLP and speech directions. On both reviewed 2026 papers he sits in a middle author slot without the corresponding role, indicating steady team participation rather than leading an independent group so far.
Activity77/100
Scores 77/100: indexed output climbs from 1 work in 2023 to 13 works in 2026, matching the recent three-year count of 13 works; two 2026 arXiv benchmark papers are already in the Evolve catalog, marking a steep recent ramp in submission pace.
Reproducibility38/100
Scores 38/100, the reproducibility mean across 2 reviewed papers. BabelArena is rated low readiness, with no significance tests, circular LLM judges, and no code release, while MuLA-Bench reaches medium with better-documented construction; releasing code and adding statistical checks would lift this axis.
Research summary
Baosong Yang is a researcher affiliated with Alibaba Group (China) whose profile, name-resolved from public OpenAlex bibliographic data, centers on multilingual evaluation of large language models, spanning both text-based LLM agents and audio-language models. The Evolve catalog holds two reviewed 2026 arXiv papers, both with him in a middle author position: BabelArena, a 16k-instance benchmark across 23 languages for LLM agents built via a verified translation pipeline, and MuLA-Bench, a 16-language long-form audio understanding benchmark constructed through multi-tier auditing. Evolve's verdicts credit both with substantial engineering scale while flagging methodology weaknesses: circular LLM judges and no code release for BabelArena (reproducibility readiness low), and LLM-constructed gold standards that complicate cross-model comparison for MuLA-Bench (medium). The OpenAlex snapshot shows a steep activity ramp, from 1 indexed work in 2023 to 13 works in 2026, with h-index 1 and i10-index 0, consistent with an early-stage or partially disambiguated author record. Collaboration is broad for this stage: 16 distinct coauthors and 12 recurring partners, including Zeyu Yang, Xinyu Zhang, Pei Zhang, and Satoshi Nakamura. Snapshot topic weights (Topic Modeling 14, NLP techniques 11, few-shot adaptation 5, multimodal 4) are broader than the reviewed evaluation work, suggesting the indexed record may aggregate more than one research line.
Collaboration pattern
The snapshot records 16 distinct coauthors and a stable core of 12 recurring partners — Zeyu Yang, Xinyu Zhang, Zibo Bi, Pei Zhang, Xize Cheng, Jin Xu, Satoshi Nakamura, Peng Kuang, Yuchun Fan, Jiangnan Li, Minghao Wu, and Jialong Tang — spanning NLP and speech directions. On both reviewed 2026 benchmarks he contributes as a middle author rather than corresponding author, consistent with a steady bench contributor inside large evaluation teams (collaboration score 56/100).
Research keywords
Highlights
BabelArena: 16k instances across 23 languages for LLM agents
MuLA-Bench: multi-tier auditing for 16-language long-form audio
Reproducibility gap: no code release for BabelArena
Timeline
- 2023First indexed OpenAlex output: 1 work recorded in this year's counts, the earliest signal in the snapshot.
- 2026Output surge: 13 works in 2026, matching the recent three-year count of 13 works.
- 2026Two multilingual evaluation benchmarks enter the Evolve catalog as middle-author contributions: BabelArena (LLM agents, 23 languages) and MuLA-Bench (long-form audio, 16 languages).
- 2026Coauthor network reaches 16 distinct collaborators, with 12 recurring partners spanning NLP and speech teams.
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 figures come from an OpenAlex author snapshot only (identity: openalex); h-index (1) and i10-index (0) may undercount if 'Baosong Yang' records are split across author IDs.
- Only 2 papers are in the Evolve-reviewed catalog, both 2026 arXiv preprints with middle-author positions; no peer-reviewed venue data is included.
- Snapshot topic counts (e.g., Topic Modeling 14, NLP techniques 11) are broader than the reviewed evaluation work, so the indexed record may mix research lines or namesakes.
- Affiliation and country ('Alibaba Group (China)', country code 'KY') are quoted as given in the source record; no ORCID is available to confirm identity.