Vineeth N. Balasubramanian
Microsoft Research (India) · United States
Microsoft Research (India) researcher in model merging and LLM code-generation evaluation; sharp 2026 output surge
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
1
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-09) 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—
Output metrics are not yet available: the underlying data does not include a reliable publication-count or venue record to score this axis. It would measure total works, venue tiers, and first/last-author share across the researcher's career.
Impact—
Impact cannot be scored yet: the record's citation evidence is too thin (h-index 1, i10-index 1). This axis would measure citation counts, h-index trajectory, and uptake of the researcher's papers by other work.
Focus0/100
This snapshot scores focus at 0: the field distribution splits exactly evenly (0.5/0.5) between LLM code-generation evaluation and model merging, and the five OpenAlex topic clusters — advanced neural network applications (13), multimodal ML applications (13), topic modeling (9), domain adaptation and few-shot learning (7), and XAI (4) — reinforce breadth over a single concentration.
Collaboration22/100
The collaboration score is 22, reflecting a narrow network: only 2 distinct co-authors appear in the OpenAlex record, with Aditya Pola and Arkaprava Majumdar as the recurring partners. Both catalog papers place this researcher in the last author position (non-corresponding), consistent with a senior supervisory role on a small stable team rather than broad multi-group collaboration.
Activity93/100
Activity scores 93 on the strength of recent output: 23 works in the last three years, with 22 recorded in 2026 alone versus 1 in 2025 and 2 in 2023. That concentration signals an intense current publishing push — both Evolve-reviewed papers are 2026 arXiv preprints.
Reproducibility55/100
Reproducibility scores 55, averaged over 2 Evolve-reviewed papers, each rated medium readiness. Evolve's verdicts flag concrete gaps: Merge++ reports broad merge gains without variance, cost, or loss-ablation reporting, and ISA-Bench's novel REG analysis rests on an unvalidated GPT-5.5 judge — careful artifacts, but incomplete verification trails.
Research summary
Vineeth N. Balasubramanian is profiled here from public bibliographic data (an OpenAlex author record; no ORCID provided), affiliated in this snapshot with Microsoft Research (India). The record's recent activity is striking: 23 works over the past three years, 22 of them logged in 2026, against 1 in 2025 and 2 in 2023. The Evolve-reviewed portfolio maps onto the two fronts that split the record's field distribution evenly (0.5/0.5): machine learning for LLM code-generation evaluation and for model merging. Merge++ (2026) proposes data-free checkpoint inversion and distillation to refine merged models, while ISA-Bench (2026) introduces 92 tasks across five instruction set architectures to probe LLM computational reasoning; in both, this researcher is the last-listed, non-corresponding author, with Aditya Pola and Arkaprava Majumdar as recurring co-authors and only 2 distinct co-authors overall. OpenAlex topic tags — advanced neural network applications (13), multimodal ML (13), topic modeling (9), domain adaptation and few-shot learning (7), and XAI (4) — indicate breadth. Impact indicators are thin (h-index 1, i10-index 1), so author disambiguation and citation metrics warrant caution.
Collaboration pattern
The collaboration footprint in this snapshot is narrow: 2 distinct co-authors in the OpenAlex record, yielding a collaboration score of 22. The recurring partners are Aditya Pola and Arkaprava Majumdar. In both catalog papers the researcher occupies the last author position (non-corresponding), a pattern consistent with senior supervision of a small, stable team rather than broad multi-group collaboration.
Research keywords
Highlights
Merge++: refining model merges with data-free checkpoint inversion
ISA-Bench: probing LLM computational reasoning across five ISAs
A two-front 2026 publishing surge
The OpenAlex record logs 22 works in 2026 versus 1 in 2025 and 2 in 2023, with the field split evenly (0.5/0.5) between LLM code-generation evaluation and model merging — the two fronts covered by the catalog's two medium-reproducibility reviews.
Timeline
- 2023Earliest year in the provided counts-by-year snapshot: the OpenAlex record logs 2 works.
- 2025The record logs 1 work in this year, the low point before the 2026 surge.
- 2026Peak output: 22 works recorded, including the two Evolve-reviewed arXiv preprints (Merge++ and ISA-Bench) with this researcher as last-listed author.
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.
- All bibliometrics come from a single OpenAlex author record (identity: openalex); no ORCID or cross-source verification is available.
- Impact indicators are unusually low (h-index 1, i10-index 1); given the relatively common name, this may reflect author disambiguation, so quantitative metrics should be read with caution.
- The qualitative assessment rests on only 2 Evolve-reviewed papers (both 2026 arXiv preprints, medium reproducibility readiness); no peer-reviewed venue data is provided.
- Affiliation metadata is unresolved: the record lists country as 'US' while the institution given is Microsoft Research (India).