Manasi Patwardhan
Tennessee Cancer Specialists · United States
NLP researcher building LLM-driven scientific ideation, evaluation benchmarks, and peer-review automation
4
papers here
4
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
0
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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 cannot be scored yet: OpenAlex records no citation-indexed publications for this identity (h-index and i10-index are both 0). This axis would measure the volume and consistency of published works once they are reliably linked to the author record.
Impact—
Impact is not scored: with h-index 0 and i10-index 0 in OpenAlex, there is not yet enough citation data to evaluate influence. This axis would measure citation uptake, scholarly reach, and downstream use of the researcher's work.
Focus0/100
The focus score is 0, reflecting an evenly spread field distribution: four sub-areas each hold 25% — NLP for AI-driven science, AI evaluation/scientometrics, LLM post-training, and LLM alignment for scientific ideation. The OpenAlex topic profile is similarly dispersed, with Topic Modeling (11 records) ahead of AI applications (4), software engineering research (3), information retrieval (3), and scientific data management (3). Output is temporally concentrated in 2026 but thematically broad rather than clustered in one niche.
Collaboration54/100
The collaboration score is 54, a mid-range value. OpenAlex links 14 distinct co-authors across the 14 works of the recent three-year window, indicating steady but not extensive co-authoring. The reviewed papers place the researcher as a middle author on three of four works and last author on one, consistent with contributing within larger teams. Recurring names in the co-author network include Arman Cohan and Yilun Zhao, the latter appearing most often in the frequent-coauthor list.
Activity79/100
Activity scores 79, driven by the recent publication window: OpenAlex counts 14 works in 2026, matching the recent3yWorks figure of 14, so all recorded output falls in a single year. Evolve's catalog independently confirms the pace with four 2026 arXiv papers spanning ideation alignment, evaluation benchmarks, and peer-review post-training. This is a high-velocity, front-loaded record; longer-run consistency cannot be assessed because countsByYear contains no earlier years.
Reproducibility55/100
Reproducibility scores 55, computed from Evolve's review of four papers (reproMean 55 over reproPapers 4). Every catalogued paper carries a medium repro-readiness rating: each proposes benchmarks or RL training pipelines whose claims depend on LLM judges or reward models — citation-trained rewards in Learning to Ideate, LLM-as-judge scoring in BackTrend and IdeaAMBIG, and a GPT-5.4 judge embedded in ActReview's own pipeline. Medium readiness signals reusable assets paired with still-unvalidated scoring machinery.
Research summary
Manasi Patwardhan is a researcher whose profile is name-resolved from OpenAlex bibliographic data, with a listed US affiliation at Tennessee Cancer Specialists. The documented output is concentrated in natural language processing applied to science: four 2026 arXiv papers reviewed by Evolve span LLM alignment for scientific ideation, AI evaluation and scientometrics, NLP for science, and LLM post-training. The work is benchmark- and method-heavy: Learning to Ideate for Scientific Impact trains reward models with reinforcement learning (GRPO) against citation signals; BackTrend provides a retrospective benchmark for weak-signal prediction in science; IdeaAMBIG measures implementation-critical gaps in research-idea specifications; and ActReview post-trains Qwen3-8B with rebuttal-guided data and rubric rewards for actionable peer review. OpenAlex records 14 works in the recent three-year window (14 in 2026 alone) and 14 distinct co-authors, supporting a high activity score (79) and a mid-range collaboration score (54). Topic Modeling dominates the topic profile (11 records) alongside AI applications, software engineering research, information retrieval, and scientific data management. Citation-based output and impact metrics (h-index, i10-index) are not yet recorded, so those axes are left unscored. All four catalogued papers carry medium reproducibility readiness, yielding a reproducibility score of 55.
Collaboration pattern
Collaboration is mid-range (score 54): 14 distinct co-authors across 14 recent works. The reviewed papers place the researcher as a middle author on three and last author on one, suggesting contribution within larger NLP-focused teams. Recurring co-authors include Yilun Zhao (listed three times), Arman Cohan, Yiling Ma, Sihong Wu, Xiao Zhou, Owen Jiang, Tiansheng Hu, Cai Xu, and Ziyu Chen. The list also contains the profile's own name, indicating deduplication noise in the source data.
Research keywords
Highlights
Reinforcement learning against citation signals to steer scientific ideation
Retrospective benchmark for weak-signal prediction in science
Quantifying implementation-critical gaps in research-idea specs
Rebuttal-guided post-training for actionable peer review
Timeline
- 2026Publishes (as last author) Learning to Ideate for Scientific Impact, aligning LLM ideation with citation-trained reward models and GRPO.
- 2026Co-authors BackTrend, a validated retrospective benchmark for scientific weak-signal prediction.
- 2026Co-authors IdeaAMBIG, benchmarking implementation-critical gaps in research-idea specifications.
- 2026Co-authors ActReview, introducing rebuttal-guided training data and rubric rewards for actionable peer review generation.
Papers in the Evolve catalog
Learning to Ideate for Scientific Impact
BackTrend: Evaluating Scientific Weak-Signal Prediction via Backward Reconstruction
IdeaAMBIG: Benchmarking Implementation-Critical Gaps in Research-Idea Specifications
ActReview: Rebuttal-Guided Training Data and Rubric Rewards for Actionable Peer Review Generation
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 come from OpenAlex only; h-index and i10-index are recorded as 0, so output and impact dimensions are unscored rather than estimated.
- Identity is name-resolved from public bibliographic data; the listed affiliation (Tennessee Cancer Specialists, US) sits oddly beside an arXiv NLP record, so affiliation and identity linking should be treated with caution.
- All four reviewed papers are 2026 arXiv preprints assessed through Evolve's review verdicts, not peer-reviewed publication records.
- The frequent-coauthor list includes the profile's own name, suggesting author-deduplication noise in the underlying data.