Navonil Majumder

Singapore University of Technology and Design · Singapore

Unclaimed · AI-structured profileORCID OpenAlex

NLP and multimodal ML researcher at SUTD extending into vision-language-action models and agent benchmarks

papers here

reviewed by Evolve

first-author

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 scored: deduplicated work counts in this OpenAlex snapshot are too incomplete to compute a reliable productivity axis. Once coverage improves, this dimension would measure annual and cumulative publication volume normalized against field norms.

Impact—

Not scored: the citation record behind this profile is insufficient for a composite impact axis, although an h-index of 29 and an i10-index of 40 are already on file. Once complete, it would measure citation influence and scholarly reach.

Focus100/100

The focus score of 100 reflects a tightly concentrated record: topic modeling leads with 43 works, followed by sentiment analysis and opinion mining (29), NLP techniques (24), multimodal machine learning applications (22), and speech and dialogue systems (22) — five adjacent topics within one language-and-multimodality cluster. The single fieldDist entry, robot learning / vision-language-action models, marks a current expansion outward from that core rather than dispersion.

Collaboration48/100

The collaboration score of 48 sits at mid-range: the snapshot records 10 distinct recent coauthors against 23 works over the last three years, a pattern of small, repeat-team projects rather than large rotating consortia. The frequent-coauthor list is anchored by Soujanya Poria, Nancy F. Chen, and Zhengyuan Liu, with Tej Deep Pala, Jianfei Yang, Vernon Toh, and others recurring; on both 2026 catalog papers he holds a middle authorship position.

Activity93/100

Activity scores 93 on sustained recent output: 23 works in the last three years, with yearly counts of 7 (2024), 7 (2025), and 9 so far in 2026, against a career peak of 17 in 2021 — the current cadence nearly matches the best year on record. Earlier counts of 2 (2017), 8 (2019), and 14 (2023) trace a steadily compounding curve after modest beginnings.

Reproducibility55/100

Reproducibility scores 55, computed by Evolve from a single reviewed paper (reproPapers 1, mean readiness 55), so the evidence base is thin but diagnostic. MNIST-PRO rates high readiness even though its n=100 cells lack variance estimates; MemBodied rates medium — its gains on RMBench and real robots are large but single-seed with n=50. Reported effects look credible; statistical rigor is the gap.

Research summary

Navonil Majumder is a researcher at the Singapore University of Technology and Design whose bibliometric record, resolved from OpenAlex, centers on natural language processing and multimodal machine learning. His topic profile is led by topic modeling (43 works), followed by sentiment analysis and opinion mining (29), NLP techniques (24), multimodal machine learning applications (22), and speech and dialogue systems (22). The snapshot records an h-index of 29 and an i10-index of 40, with 23 works in the last three years and 9 works already counted in 2026 — a cadence approaching his 2021 peak of 17. Recent cataloged work extends this language-centric core into embodied AI: MemBodied (2026) adds recurrent associative memory to vision-language-action models, and MNIST-PRO (2026) recasts MNIST as a glimpse-based partially observable environment for multimodal LLM agents. Evolve reviews credit both papers with coherent, diagnostic findings while flagging single-seed results and missing variance estimates. He works within a compact Singapore-based network of 10 distinct recent coauthors, including Soujanya Poria, Nancy F. Chen, and Zhengyuan Liu, and holds middle authorship on both 2026 catalog papers. This profile is name-resolved from public bibliographic data.

Collaboration pattern

The pattern is a compact, repeat-team Singapore network: 10 distinct recent coauthors across 23 works in three years. Soujanya Poria and Nancy F. Chen anchor the frequent-coauthor list, with Zhengyuan Liu, Tej Deep Pala, Vernon Toh, Bryce Goh, Raphael Yee, Jianfei Yang, and Liming Chen also recurring. On both 2026 catalog papers he holds a middle authorship position, and the mid-range collaboration score of 48 is consistent with this small-team, specialist-contributor style rather than large consortium work.

Research keywords

Natural language processingTopic modelingSentiment analysis and opinion miningMultimodal machine learningSpeech and dialogue systemsVision-language-action modelsRobot manipulationAssociative and episodic memoryMultimodal LLM agentsActive perceptionPartially observable MDPs

Highlights

Recurrent associative memory for vision-language-action models
MemBodied (2026) adds fixed-size associative episodic memory to VLAs using a gated delta rule; Evolve finds the gains on RMBench and real robots large, but notes single-seed, n=50 results limit rigor. Reproducibility readiness: medium.
MNIST recast as a partially observable world for AI agents
MNIST-PRO (2026) turns MNIST into glimpse-based partially observable search for multimodal LLM agents, probing active perception and working memory; Evolve calls the findings coherent and diagnostic, while n=100 cells lack variance estimates. Readiness: high.
From language-centric NLP to embodied agents
The 2026 catalog papers extend a core that bibliometrics place in topic modeling (43 works), sentiment analysis (29), and multimodal ML (22) into vision-language-action models and agent benchmarks — a pivot visible in the single fieldDist entry for robot learning / VLAs.

Timeline

2017Earliest year in the bibliometric record, with 2 works — the starting point of a steadily compounding output curve.
2021Career peak on record: 17 works, more than double the 8 recorded in 2020.
2023Output re-accelerates to 14 works after 10 in 2022.
20257 works recorded, level with 2024 — a stable recent baseline within the last-three-year total of 23.
20269 works counted so far; the Evolve catalog includes two of them — MemBodied and MNIST-PRO — both extending the record into embodied and agent research.

Papers in the Evolve catalog

Frequent co-authors

Vernon Toh×1 Zhengyuan Liu×1 Nancy F. Chen×1 Soujanya Poria×1 Tej Deep Pala×1 Bryce Goh×1 Raphael Yee×1 Jianfei Yang×1 Liming Chen×1 Soujanya Poria×1

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.

Bibliometrics (h-index, i10-index, topic and per-year counts) come from an OpenAlex snapshot; the profile is name-resolved from public bibliographic data.
The output and impact dimensions are unscored (dimMissing): deduplicated work and citation coverage in the snapshot is incomplete.
The reproducibility score rests on a single reviewed paper, so the mean of 55 is not a robust estimate.
Only two papers carry full Evolve reviews, and both are 2026 arXiv preprints that may be superseded by peer-reviewed versions.