Bo Zheng: verification-centric AI research across factuality evaluation, LLM reasoning, and AIGC detection
3
papers here
3
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
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—
Not yet scored: this profile lacks publication-volume data such as total paper counts, venue types, and yearly productivity. The output axis would measure how many peer-reviewed works Zheng produces and where they appear.
Impact—
Not yet scored: no citation counts, h-index, or usage signals are available in the underlying record. The impact axis would measure how often Zheng's papers are cited and the influence of the venues publishing them.
Focus0/100
The focus score is 0, reflecting a field distribution split exactly into thirds: 33.3% natural language processing (factuality evaluation / LLM agents), 33.3% machine learning (LLM reasoning), and 33.3% multimedia forensics / computer vision (AIGC detection). No single field dominates the reviewed record, so the low score describes breadth across three research communities rather than an assessment of work quality.
Collaboration74/100
The collaboration score is 74, grounded in 40 distinct coauthors appearing across Zheng's bibliographic records, with 12 recurring collaborators such as Kening Zheng, Yazhi Guo, and Wei-Chieh Huang. This indicates broad, team-based work rather than solo research, though all three reviewed papers are 2026 arXiv preprints and no corresponding-author role is recorded for Zheng in them.
Activity—
Not yet scored: the longitudinal publication-frequency data needed to gauge recent research velocity is unavailable. The activity axis would measure year-over-year output tempo and whether Zheng remains actively publishing.
Reproducibility43/100
The reproducibility score averages 43 across three reviewed papers. Evolve rates PIVOT as low readiness, citing a closed LLM stack and no significance tests for its 70-72% detection accuracy; EAVer and the PMPS reasoning paper are medium readiness, with EAVer lacking code release and variance reporting and PMPS showing single-run, math-only gains. Consistent statistical rigor and artifact release would raise this axis.
Research summary
Bo Zheng is an AI researcher whose reviewed work splits evenly across three areas: natural language processing (factuality evaluation / LLM agents), machine learning (LLM reasoning), and multimedia forensics (AIGC detection). Evolve has reviewed three 2026 arXiv papers involving Zheng. As first author of PIVOT, Zheng introduced physics-grounded verification that detects AI-generated audio-video at 70-72% accuracy versus 54-57% for Gemini-based inspection. EAVer, a middle-author contribution, trains an 8B agentic verifier on privileged-teacher trajectories, cutting searches by roughly 80% while reaching top Macro-F1 on long-form factuality verification. The PMPS paper adds prototype-based process supervision to latent chain-of-thought reasoning, consistently beating SIM-CoT and CODI. A recurring theme is verification — of physical consistency in synthetic media, of claims in long text, and of intermediate reasoning steps. Bibliometrics show a collaboration score of 74 (40 distinct coauthors, 12 recurring) against a focus score of 0 from an even three-way field split, and a reproducibility mean of 43, weighed down by closed LLM stacks, absent significance tests, and missing code or variance reporting. Institution and country are not recorded; as 'Bo Zheng' is a common name, this profile is name-resolved from public bibliographic data.
Collaboration pattern
With 40 distinct coauthors and a collaboration score of 74, Zheng works in sizeable teams. Twelve collaborators recur across records, including Kening Zheng, Aoying Zheng, Zhigang Chang, Yazhi Guo, and Miaotian Guo, alongside Qingwei Zong, Xianhai Xie, Weiqiang Jin, Chengze Li, Hanrong Zhang, Jie Yang, and Wei-Chieh Huang. In the reviewed papers Zheng is first author once and middle author twice, with no corresponding-author role — an early-career, team-embedded pattern.
Research keywords
Highlights
Physics-law verification for detecting AI-generated audio-video
Efficient agentic verifier for long-form factuality
Prototype-based process supervision for latent chain-of-thought
Timeline
- 2026First-authors PIVOT, introducing physics-grounded verification that detects AI-generated audio-video at 70-72% accuracy versus 54-57% for Gemini-based inspection.
- 2026Contributes to EAVer, which frames long-form factuality verification as an end-to-end agentic policy trained with privileged-teacher trajectories, achieving roughly 80% fewer searches and top Macro-F1.
- 2026Contributes to the PMPS work on structural process supervision for latent chain-of-thought reasoning, consistently outperforming SIM-CoT and CODI.
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 indicators are derived from public bibliographic data (OpenAlex-based) and Evolve's reviews of three arXiv preprints; no citation counts or h-index are available.
- 'Bo Zheng' is a common name; identity is name-resolved from public bibliographic records, and institution, country, and ORCID are not recorded.
- All reviewed papers are 2026 arXiv preprints; peer-review status and venue acceptance are not reflected in this profile.
- The aggregate of 40 distinct coauthors may include homonymous authors and spans the full bibliographic record, not only the three reviewed papers.