Microsoft Research researcher pairing a long NLP record (h-index 84) with LLM alignment and multi-agent systems
3
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
84
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: the underlying data for this axis (total publication volume, per-year and venue-normalized productivity) has not been computed in this snapshot. It would measure how many works the researcher produces and how that output is distributed across venues and years.
Impact—
Not yet scored: citation-based impact indicators (total and field-weighted citations, citation trajectories) are not available in this snapshot. This axis would measure the influence of the researcher's work on the subsequent literature, beyond the raw h-index and i10-index counts already on file.
Focus0/100
The focus dimension scores 0, reflecting a wide topical spread rather than a narrow specialty: OpenAlex topics run from Topic Modeling (389 works) and NLP techniques (342) to multimodal learning (150), speech recognition and synthesis (69) and text analysis (68), while the current field distribution splits evenly between LLM alignment and multi-agent systems.
Collaboration64/100
Collaboration scores 64 on the 0-100 scale, built on 24 distinct co-authors in the recent three-year window and a recurring cluster of twelve frequent collaborators including Peng Lai, Li Dong, Shaohan Huang, Dongdong Zhang and Peng Li. In the reviewed papers the researcher appears as last author twice and middle author once, consistent with a senior supervisory position within sizeable teams.
Activity100/100
Activity is scored at the maximum of 100. The record shows 154 works in the most recent three-year window and an annual trajectory that climbed from single digits in the mid-2000s to 92 works in 2022, staying high at 80 (2023), 81 (2024), 42 (2025) and 31 recorded so far in 2026.
Reproducibility55/100
Reproducibility scores 55 (reproMean 55 across the 2 papers with completed reviews). Both reviewed papers are rated medium readiness: Agensh reports no variance, cost or orchestrator baseline, and AlignDiff reports no variance or significance tests — a consistent pattern of strong reported results with limited statistical reporting.
Research summary
Furu Wei is a researcher at Microsoft Research whose profile, resolved from OpenAlex bibliographic data, shows a long-standing natural language processing career that has pivoted toward large language model alignment and LLM multi-agent systems. The snapshot reports an h-index of 84 and an i10-index of 293. Annual output rose from single digits in the mid-2000s to a 2022 peak of 92 works, and 154 works appear in the most recent three-year window, earning the maximum activity score of 100. Topic coverage spans Topic Modeling (389 works), NLP techniques (342), multimodal machine learning (150), speech recognition and synthesis (69) and text analysis (68); the current field distribution splits evenly, 50/50, between LLM alignment and agentic AI. Evolve-reviewed work includes Agensh, which scales an orchestrator-free harness to 1,024 agents with a reported +49% relative gain at 128 agents on ProgramBench, and AlignDiff, a two-stage DPO preference-data selection method built on reward margins and NLL-gap difficulty. Both are rated medium reproducibility (reproMean 55 over 2 papers). Collaboration involves 24 distinct recent co-authors and a stable cluster including Li Dong, Shaohan Huang and Dongdong Zhang.
Collaboration pattern
The researcher works in sizeable teams: 24 distinct co-authors appear in the recent three-year window, with a recurring cluster of twelve frequent collaborators including Li Dong, Shaohan Huang, Dongdong Zhang, Peng Li, Yang Liu and Guanhua Chen. Positioning in reviewed papers — last author on Agensh and the multimodal geometry paper, middle author on AlignDiff — suggests a senior supervisory role on agentic and multimodal projects and a contributing role on alignment work.
Research keywords
Highlights
Scaling an orchestrator-free harness to 1,024 LLM agents
Model-intrinsic signals guide DPO preference data selection
Credit-addressable reasoning for multimodal geometry
Listed in the catalog as a 2026 arXiv paper with the researcher in the last author position; it has not yet received a full Evolve review, so no verdict, keywords or reproducibility rating are available.
Timeline
- 2026Published Agensh as last author, scaling an orchestrator-free multi-agent harness to 1,024 agents (arXiv).
- 2026Published AlignDiff as a middle author, on model-intrinsic preference data selection for DPO alignment (arXiv).
- 2026Appeared as last author on a credit-addressable reasoning paper for multimodal geometry (arXiv); Evolve review pending.
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 (h-index, i10-index, yearly counts, topics) come from an OpenAlex-derived snapshot; the output and impact dimensions lack computed scores.
- Qualitative review coverage is limited to three 2026 arXiv papers in the Evolve catalog; one of them lacks a verdict, keywords and a reproducibility rating.
- Affiliation metadata is reported as provided — Microsoft Research (United Kingdom) with country listed as 'US' — an inconsistency passed through unverified.
- Identity is name-resolved from public OpenAlex bibliographic data via ORCID 0000-0002-7810-5852; biographical information beyond these records is not covered.