University of Hong Kong researcher in LLM reasoning: latent steering and self-improving reasoning models.
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-17) 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: publication-volume data is insufficient in the current snapshot. This axis would measure the number, venue spread and consistency of the researcher's outputs over time. No count is estimated here.
Impact—
Not yet scored: citation and influence data is unavailable, so no h-index or citation counts are quoted. This axis would measure how much the two 2026 preprints and other outputs are cited and built upon.
Focus100/100
Scores 100 — the top of the scale. The field distribution contains exactly one area, natural language processing (LLM reasoning), and both Evolve-reviewed 2026 papers stay inside it: a training-free latent-steering method for efficient reasoning and a survey of scaling large reasoning models. No catalog work strays from this theme, so topical concentration is maximal on current evidence.
Collaboration68/100
Scores 68. The bibliometric record lists 30 distinct coauthors, and 12 recur frequently, including Hengyuan Zhang, Chenming Shang, Kailai Yang, Hayden Kwok-Hay So and Ngai Wong. Both catalog papers carry middle authorship with no corresponding role, which fits a steady contributor inside a multi-author LLM reasoning group rather than a solo research line; the broad network still leaves room to deepen key partnerships.
Activity—
Not yet scored: the snapshot lacks a longitudinal publication time series, so recent-year cadence cannot be computed. This axis would measure publishing pace and continuity over recent years; only the two 2026 papers are currently observable.
Reproducibility55/100
Scores 55, computed as the mean over one reviewed paper (reproPapers: 1), so the evidence base is narrow. The SPS latent-steering paper rates medium: it is training-free and reports consistent Pass@4 gains on Qwen3 1.7B-14B across 9 benchmarks, but statistical rigor is thin. The self-improvement survey rates low, offering a reward taxonomy rather than experiments.
Research summary
Jing Xiong is a researcher at the University of Hong Kong working on large language model reasoning. The Evolve catalog holds two 2026 arXiv papers, both with middle authorship and no corresponding role. The first, Efficient Reasoning Exploration via State-Conditioned Latent Steering with Progress Guidance, introduces SPS, a training-free latent-steering method with progress guidance; Evolve's review reports consistent Pass@4 gains on Qwen3 models from 1.7B to 14B across 9 benchmarks, while flagging thin statistical rigor and medium reproducibility readiness. The second, Scaling Large Reasoning Models beyond Human Supervision, is a survey that organizes LRM self-improvement along two axes with an L0-L4 ladder and a thorough reward taxonomy; it contains no original experiments, its title overclaims, and reproducibility readiness is low. Available dimensions score focus at 100 (a single field, natural language processing (LLM reasoning)), collaboration at 68 (30 distinct coauthors, 12 recurring), and reproducibility at 55 averaged over one paper; output, impact and activity lack sufficient data. Because 'Jing Xiong' is a common name, this profile is name-resolved from public bibliographic data.
Collaboration pattern
Collaboration scores 68. The record shows 30 distinct coauthors, with 12 recurring frequently — Hengyuan Zhang, Chenming Shang, Zunhai Su, Xiao Liang, Hui Shen, Dawei Li, Shiping Yang, Kailai Yang, Wei Zhang, Ruobing Xie, Hayden Kwok-Hay So and Ngai Wong. In both catalog papers Jing Xiong is a middle author and not corresponding, consistent with contributing within a multi-author LLM reasoning team rather than leading an independent research line.
Research keywords
Highlights
Training-free latent steering (SPS) for efficient LLM reasoning
Survey charting self-improving reasoning beyond human supervision
Timeline
- 2026Co-authored (middle authorship) the arXiv paper 'Efficient Reasoning Exploration via State-Conditioned Latent Steering with Progress Guidance', introducing the training-free SPS method for LLM reasoning.
- 2026Contributed to the arXiv survey 'Scaling Large Reasoning Models beyond Human Supervision: A Path toward Superintelligence', covering the L0-L4 ladder and a thorough reward taxonomy.
- 2026Both catalog outputs fall within LLM reasoning / large reasoning models, yielding a single-field focus score of 100 alongside a network of 30 distinct coauthors in the same year.
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.
- Name-resolved from public bibliographic data: 'Jing Xiong' is a common name, and no ORCID or country metadata was provided, so identity is not independently verified.
- Reproducibility (55) is a mean over a single reviewed paper (reproPapers: 1) and may not generalize.
- Output, impact and activity dimensions are unscored because the underlying bibliometric data is insufficient; no citation counts or publication volumes were provided.
- The catalog covers only two Evolve-reviewed 2026 arXiv preprints; other publications, if any, are not reflected here.