Alibaba Group NLP researcher advancing LLM reasoning training, text reranking, and agentic coding environments
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
43
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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—
Evolve has not computed an output score for this profile: the axis would measure publication volume and consistency over time, but the snapshot lacks the complete verified works record needed to score it, despite 72 recent works being recorded.
Impact—
No impact score is available yet. This axis would quantify citation-based influence (for example citations per work); the snapshot lists an h-index of 43 and an i10-index of 128, but not the full citation record Evolve needs to score this dimension.
Focus0/100
Focus scores 0/100 because the field distribution is perfectly even: NLP for LLM reasoning training, information retrieval/NLP, and machine learning for code/LLM agents each account for exactly one-third of the profile. Topical records are similarly spread across topic modeling (187), NLP techniques (139), multimodal machine learning (88), speech and dialogue (35), and domain adaptation/few-shot learning (32), indicating breadth rather than concentration.
Collaboration70/100
Collaboration scores 70/100 on the basis of 33 distinct coauthors in the snapshot. On the three reviewed papers Huang appears as last author on the reranking work and as a middle author on the other two, never as corresponding author, consistent with a senior contributor within sizeable Alibaba teams; recurring partners include Dayiheng Liu, Xingzhang Ren, Tianyi Tang, Kexin Yang, and Jinyang Zhang.
Activity100/100
Activity scores 100/100, the maximum, reflecting an intense recent cadence: 72 works in the last three years, with 40 works in 2023, 39 in 2024, and 29 in 2025 per the counts-by-year record, on top of a 2021 peak of 53. All three Evolve-reviewed papers appeared in 2026.
Reproducibility32/100
Reproducibility scores 32/100, averaging Evolve's readiness ratings across the three reviewed papers (low, medium, low). Reviews cite no code release for Terminal-Universe, duplicated baselines and no significance tests in the reasoning-data paper, and efficiency claims needing scrutiny in the reranking work.
Research summary
Fei Huang is a researcher at Alibaba Group whose recent work centers on training and evaluating large language models for reasoning, retrieval, and agentic coding. The bibliometric snapshot (OpenAlex identity, ORCID 0000-0002-3709-5053) attributes an h-index of 43 and an i10-index of 128, with 72 works in the last three years and a recorded peak of 53 works in 2021. Topical coverage is broad: topic modeling (187 records), natural language processing techniques (139), multimodal machine learning applications (88), speech and dialogue systems (35), and domain adaptation/few-shot learning (32). Evolve has reviewed three 2026 arXiv preprints spanning the profile's three evenly weighted fields — multi-perspective evidence reranking (last author), code-centric reasoning data synthesis, and Terminal-Universe agent environments. Reviewers credited strong reported results (e.g., +11.9 on TB2.1) while flagging low reproducibility readiness: missing code releases, duplicated baselines, and absent significance tests, yielding a reproducibility mean of 32/100. Activity scores 100/100 and collaboration 70/100 (33 distinct coauthors), but the evenly split field distribution produces a focus score of 0, and the output and impact axes remain unscored.
Collaboration pattern
The snapshot records 33 distinct coauthors (collaboration dimension 70/100). Recurring partners center on an Alibaba NLP cluster: Dayiheng Liu, Xingzhang Ren, Tianyi Tang, Kexin Yang, and Jinyang Zhang, plus Weibin Liao, Keqin Bao, Sihang Li, Shaobo Wang, Muyang Ye, Hongxin Ding, Yue Fang, and Hongxin Ding's peers. On the three reviewed papers Huang appears once as last author and twice as a middle author, never as corresponding author.
Research keywords
Highlights
Staged RL and multi-perspective reasoning for LLM reranking
Code-centric synthesis turns programs into reasoning data
Terminal-Universe recycles agent trajectories into environments
Timeline
- 2002Earliest publication year recorded in the OpenAlex counts-by-year data (1 work); given the common name, early coverage may reflect namesakes.
- 2020Recorded annual output jumps to 23 works, up from 5 in 2019, marking a step change in the counts-by-year data.
- 2021Peak year in the snapshot, with 53 recorded works.
- 2023Sustained high output: 40 works in 2023 and 39 in 2024 per the counts-by-year record, contributing to 72 works over the last three years.
- 2026Three arXiv preprints enter the Evolve catalog: multi-perspective evidence reranking, code-centric reasoning data synthesis, and Terminal-Universe agent environments.
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.
- Identity resolved from public bibliographic data (OpenAlex) plus ORCID 0000-0002-3709-5053; Fei Huang is a common name, and early records (2002-2015) in the counts-by-year data may include namesakes.
- All bibliometrics come from an OpenAlex-derived snapshot; the h-index (43) and i10-index (128) are quoted as provided and were not independently verified.
- The output and impact dimensions are unscored (insufficient radar data); no estimates are offered for them.
- The reproducibility score and verdicts rest on only three Evolve-reviewed 2026 arXiv preprints, none confirmed as peer-reviewed.