NLP and LLM-agent researcher linking simulated user mental states to multi-agent scientific hypothesis discovery
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-20) 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 data is available yet: publication counts and venue totals are absent for this name-resolved profile. This axis would measure how much work the researcher produces and where it appears.
Impact—
Impact cannot be scored because citation counts and h-index-style measures are not in the provided data. This axis would measure how often and how strongly later work builds on these papers.
Focus0/100
Focus scores 0, and the underlying field distribution explains it bluntly: this record splits exactly 50% natural language processing and 50% AI for science / LLM agents and scientific hypothesis discovery. Each 2026 catalog paper sits in a different field — Mind2Dialogue in NLP, HypoEvolve in agent-based hypothesis discovery — so the axis reads a two-front agenda rather than accumulation in one field.
Collaboration69/100
Collaboration is the strongest measured axis at 69, grounded in 31 distinct coauthors across the record. The frequent-coauthor list mixes names such as Zhen Wang, Jieyuan Liu, Mengzhou Hu, Jefferson Chen, and Hsin-Yuan Lee with Trey Ideker, Dexter Pratt, Pratibha Jagannatha, and Wei Wang, spanning both cataloged fields. Note that Zhen Wang appears twice and may be two different people.
Activity—
Activity data is insufficient: beyond the two 2026 arXiv papers, no publication-year distribution or recent-output counts exist. This axis would measure research cadence and consistency across recent years.
Reproducibility20/100
Reproducibility scores 20, computed as the mean readiness over 2 papers, and both catalog entries are rated low. Evolve's verdicts explain why: Mind2Dialogue depends on a closed Oracle teacher model, while HypoEvolve leaves its gap to the strongest baseline unresolved on one metric and is outperformed on held-out Open Targets by a constant-drug control.
Research summary
Zhiting Hu is a full-tier, name-resolved profile built from public bibliographic data, with no institution or ORCID attached. The cataloged record sits at the junction of natural language processing and AI for science through LLM agents, split exactly 50/50 between the two fields. Evolve reviews two 2026 arXiv papers, in both of which Hu holds a middle, non-corresponding authorship. Mind2Dialogue trains human-aware language models by simulating user mental states through an Oracle teacher, with personalization gains that hold across three backbones. HypoEvolve uses a generational genetic algorithm to coordinate multi-agent LLMs for scientific hypothesis discovery, lifting drug-repurposing scores over six baselines while leaving one strongest-baseline gap unresolved. Collaboration is the strongest measured dimension at 69, drawing on 31 distinct coauthors, with Zhen Wang, Jieyuan Liu, Mengzhou Hu, and Trey Ideker among recurring partners. Reproducibility is the weakest measured dimension at a mean of 20 across both papers, both rated low readiness. Output, impact, and activity metrics are unavailable, so no claims about productivity volume, citations, or cadence are made here.
Collaboration pattern
Collaboration is the record's strongest axis (69) with 31 distinct coauthors. The frequent list combines Zhen Wang (twice — likely two people), Jieyuan Liu, Mengzhou Hu, Jefferson Chen, JungHo Kong, Yiming Gao, and Hsin-Yuan Lee with Pratibha Jagannatha, Dexter Pratt, Trey Ideker, and Wei Wang, reflecting the two-field split. Both catalog papers are middle-authored and non-corresponding, so the profile reads as a contributor within larger teams rather than a solo lead.
Research keywords
Highlights
Training assistants that model the user's mental state
Genetic search over LLM agents for drug-repurposing hypotheses
Timeline
- 2026Mind2Dialogue appears on arXiv, introducing Oracle-based simulation of user mental states to train human-aware language models.
- 2026HypoEvolve appears on arXiv, applying a generational genetic algorithm over multi-agent LLMs to scientific hypothesis discovery in drug repurposing.
- 2026Both papers enter the Evolve catalog with low reproducibility-readiness ratings, and the profile's field distribution settles at an even 50/50 split between NLP and LLM-agent scientific discovery.
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.
- Profile is name-resolved from public bibliographic data; no institution, country, or ORCID was provided, and 'Zhiting Hu' is a name shared by multiple researchers.
- Bibliometrics and radar scores come from Evolve's public bibliometric snapshot only; output, impact, and activity axes lack underlying data.
- Only two 2026 arXiv preprints are in scope; peer-review status, venues, and per-paper author lists are unknown.
- 'Zhen Wang' appears twice among frequent coauthors and may denote two different individuals.