Early-career LLM agent evaluation researcher building controllable user simulators and decision-rehearsal testbeds
2
papers here
2
reviewed by Evolve
2
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—
Not yet scored. This OpenAlex record lacks a dependable cumulative publication count beyond 4 recent-3-year works, so lifetime output volume and venue mix cannot be measured. This axis would quantify total publications and sustained production over time.
Impact—
Not yet scored. OpenAlex reports h-index 0 and i10-index 0 — citations have not accumulated for this early record, which measures uptake rather than proven quality. The axis would capture citations, downstream influence, and adoption of released artifacts.
Focus0/100
Scored 0/100, the lowest focus band. OpenAlex topic assignments scatter across five clusters — topic modeling (2), speech and dialogue systems (2), natural language processing techniques (1), software testing and debugging (1), and ethics and social impacts of AI (1) — with no dominant area, and the field distribution splits evenly 0.5/0.5 between LLM agent evaluation and computational social simulation. The two reviewed papers do cohere around agent evaluation, so some dispersion may reflect name-resolution noise.
Collaboration46/100
Scored 46/100. The record lists 9 distinct coauthors over the trailing three years, a moderate-sized network; recurring names include Mayank Malhotra, Eshita, Sneha Teja Sree Reddy Thondapu, Arup Kumar Das, Jitesh Chandra Mishra, Arun Menon, Avinash Karn, and Mouli V. Both cataloged papers are first-author but not corresponding, consistent with hands-on contribution inside a team rather than leading a group.
Activity47/100
Scored 47/100. All indexed output is recent: OpenAlex counts 4 works in 2026 and 4 works across the trailing three years, and both Evolve-cataloged papers are 2026 arXiv preprints. The profile shows a concentrated recent burst of first-author output rather than a multi-year cadence; the mid-range score reflects the short observable window.
Reproducibility55/100
Scored 55/100, the mean over 2 reproducibility-assessed papers; Evolve rates both as medium readiness. Strengths include explicit artifacts — a 23-dimension persona specification backed by 64,698 conversations, and a decision lab evaluated on 50 real cases — but reviewers flag missing flat-prompt baselines, no significance tests, LLM-judge-only scoring, and small-n analysis.
Research summary
Rahul Khedar is an early-career researcher working at the intersection of large language model (LLM) agents and evaluation methodology. The Evolve catalog records two 2026 arXiv preprints, both with Khedar as first author. 'A Three-Tier Persona Vector for Controllable User Simulation in Agentic Evaluation' specifies a 23-dimensional persona schema for controlling simulated users and supports it with production-scale evidence spanning 64,698 conversations, though Evolve's review notes the absence of a flat-prompt baseline, of significance testing, and a reliance on LLM-as-a-judge scoring. 'Augur: A Synthetic Decision Lab for Rehearsing Reactions to Product and Policy Changes' builds a decision-simulation environment; across 50 real cases it argues that frontier-versus-open-weights gaps are largely prompt artifacts rather than capability differences — a claim Evolve finds credible and honest but based on a small sample. OpenAlex bibliometrics confirm an emerging profile: 4 works in the trailing three years, an h-index and i10-index of 0, and 9 distinct coauthors. Topic assignments span topic modeling, speech and dialogue systems, NLP techniques, software testing and debugging, and the ethics and social impacts of AI. The field distribution splits evenly (0.5/0.5) between LLM agent evaluation and computational social simulation. This profile is name-resolved from public bibliographic data; no institutional affiliation or country is attached to the record.
Collaboration pattern
The record shows 9 distinct coauthors across the trailing three years, a moderate-size network. Recurring collaborators include Mayank Malhotra, Eshita, Sneha Teja Sree Reddy Thondapu, Arup Kumar Das, Jitesh Chandra Mishra, Arun Menon, Avinash Karn, and Mouli V. Both Evolve-cataloged papers are first-author but not corresponding-author works, suggesting Khedar contributes hands-on technical leadership within a team while a colleague holds corresponding responsibility. The mid-range collaboration score (46) matches this pattern.
Research keywords
Highlights
Three-tier persona vector for controllable user simulation
Augur: rehearsing product and policy decisions in a synthetic decision lab
Timeline
- 2026Releases 'A Three-Tier Persona Vector for Controllable User Simulation in Agentic Evaluation' on arXiv as first author — a 23-dimension persona specification supported by evidence from 64,698 conversations.
- 2026Releases 'Augur: A Synthetic Decision Lab for Rehearsing Reactions to Product and Policy Changes' on arXiv as first author; on 50 real cases it argues frontier-versus-open gaps are mostly prompt artifacts, not capability.
- 2026OpenAlex indexes 4 works for the year — the entire recent-3-year output — split evenly between LLM agent evaluation and computational social simulation, with 9 distinct coauthors.
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 (h-index, i10-index, topics, work and coauthor counts) come from OpenAlex only; no institutional affiliation, country, or ORCID is attached, so the profile is name-resolved from public bibliographic data.
- Output and impact dimensions are unscored because indexed citations have not accumulated (h-index 0) and lifetime work counts are unavailable; scores may shift quickly as this 2026 record matures.
- Only 2 papers carry full Evolve reviews; other topic tags (topic modeling, speech and dialogue systems, software testing) come from OpenAlex clustering and may include homonymous researchers.
- No peer-reviewed venue information is present in the input; both cataloged papers are arXiv preprints rated medium for reproducibility readiness.