NLP researcher focused on LLM adaptation — fine-tuning strategies and SFT-vs-RL for tool-calling agents
3
papers here
3
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
5
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) 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 scored output dimension is available in this snapshot. This axis would measure publication volume and productivity over time; yearly counts exist, but the underlying record is insufficient to compute a comparable score.
Impact—
Impact could not be scored from the current records. This axis would capture citation uptake and scholarly influence; the snapshot lists h-index 5 and i10-index 4 but lacks fuller per-paper citation data for scoring.
Focus0/100
The focus score is 0, the lowest reading on the scale. The field distribution is split evenly — 0.5 natural language processing and 0.5 natural language processing / LLM agents — and output spans five topic clusters: Topic Modeling (25), NLP Techniques (14), Advanced Text Analysis (6), Speech Recognition and Synthesis (4), and Data Quality and Management (4). The spread sits within NLP overall, but no single cluster dominates the record.
Collaboration58/100
The collaboration score is 36 on a 0-100 scale, with only 5 distinct co-authors recorded. Both catalog papers list this researcher as last author and not corresponding, a pattern consistent with a senior contributor on small teams. Frequent co-authors Xue-Yong Fu and Md Tahmid Rahman Laskar recur in the record. The compact network points to deep, repeated partnerships rather than broad cross-institutional collaboration, though institutional data is absent.
Activity81/100
Activity scores 81, indicating high recent productivity. Yearly counts show 1 work in 2018, 1 in 2021, a peak of 8 in 2022, then 3 (2023), 4 (2024), 6 (2025), and 5 in 2026, matching the 15 works recorded over the last three years. The current run of 15 works since 2024 is the strongest sustained stretch in the record.
Reproducibility58/100
Evolve's reproducibility assessment averages 78 across the 2 reviewed papers. The SFT-versus-RL tool-calling study is rated high, reflecting an unusually thorough controlled design across six Qwen3 scales and three corpora; the customer-support fine-tuning study is rated medium, with the review noting missing significance tests and that half the data is private. The pattern indicates strong experimental rigor with room to improve data and statistical transparency.
Research summary
Shashi Bhushan TN is a natural language processing researcher whose recent work centers on how large language models should be adapted for applied settings — comparing fine-tuning strategies, parameter-efficient methods such as LoRA, and supervised-versus-reinforcement-learning training for tool-calling agents. The Evolve catalog reviews two 2026 arXiv papers on which this researcher is the last author (not corresponding): a controlled comparison of fine-tuning strategies across 13 customer-support LLMs, and an unusually thorough SFT-versus-GRPO study across six Qwen3 scales and three tool-calling corpora. Bibliometric data resolved from OpenAlex shows activity concentrated in recent years: 15 works in the last three years, with yearly counts of 4 (2024), 6 (2025), and 5 (2026), after a peak of 8 in 2022. Topic Modeling is the largest topic cluster (25), followed by Natural Language Processing Techniques (14), with smaller clusters in advanced text analysis, speech recognition and synthesis, and data quality. The listed h-index is 5 with an i10-index of 4. Evolve's reproducibility assessment averages 78 across the two reviewed papers — the tool-calling study rated high, the support-LLM study medium, reflecting private data and missing significance tests. Collaboration is compact, with five distinct co-authors recorded, including frequent collaborators Xue-Yong Fu and Md Tahmid Rahman Laskar.
Collaboration pattern
Collaboration is compact: 5 distinct co-authors are recorded, and the frequent-collaborator list centers on Xue-Yong Fu and Md Tahmid Rahman Laskar. On both catalog papers Shashi Bhushan TN appears as last author without corresponding-author status, a position typically associated with senior or supervisory contributions on small teams. The collaboration score of 36 reflects this small, repeated network; no institutional or country data is available to map affiliations.
Research keywords
Highlights
Controlled SFT-vs-RL study for tool-calling agents
Fine-tuning strategy selection for customer support LLMs
Timeline
- 2018First year with a recorded work in the snapshot (1 publication).
- 2022Most productive year in the record, with 8 works.
- 2024Output climbs again with 4 works, beginning the current three-year run.
- 20256 works recorded, the highest count since the 2022 peak.
- 20265 works recorded so far, including two arXiv studies on LLM adaptation reviewed in the Evolve catalog, both with this researcher as last 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.
- Bibliometric figures are resolved from an OpenAlex identity; institution, country, and ORCID are absent from the snapshot, so name-based resolution carries some ambiguity risk.
- Output and impact dimensions could not be scored; h-index (5) and i10-index (4) appear as raw values, but per-paper citation records are not available.
- Only the two 2026 arXiv papers carry Evolve review verdicts and reproducibility ratings; the other works counted in the yearly totals are unreviewed here.
- The reproducibility average (78) is based on just 2 reviewed papers, a small sample.