NLP and AI agent safety researcher: calibrated small-LM scam screening and agentic PDF-forgery evaluation
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
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—
Not yet scored. The underlying data on publication volume, authorship positions, and venue distribution over time is insufficient. This axis would measure total research output and productivity across the researcher's record.
Impact—
Not yet scored. Citation counts, h-index, and downstream influence metrics are unavailable in the current snapshot. This axis would measure how widely the work is cited and its influence on subsequent research.
Focus0/100
The focus score of 0 reflects an even field distribution: across the two catalogued 2026 papers, natural language processing and AI agent safety with document forensics each account for 50%. No dominant specialization has emerged yet; the work spans applied NLP systems (scam-call screening) and agentic capability evaluation (PDF forgery), showing breadth over concentration at this early stage.
Collaboration48/100
The collaboration score of 48 derives from 10 distinct co-authors across the catalogued work. On both 2026 preprints Ankit Raj sits in a middle authorship position without a corresponding-author role. Frequent collaborators include Simiao Ren, Tommy Duong, Yuxin Zhang, Dennis Ng, Xingyu Shen, Kidus Zewde, Yuchen Zhou, and Neo Tiangratanakul, indicating team-based rather than solo research output.
Activity—
Not yet scored. Longitudinal publishing cadence — papers per year and recency trends — cannot be computed from the current data. This axis would measure sustained research activity and output pace over time.
Reproducibility55/100
The reproducibility score of 55 derives from 2 scored papers, both rated medium reproducibility readiness in Evolve's review. The scam-screening preprint is candid about test-set exposure but builds entirely on synthetic data; the PDF-forgery study applies rigorous rule-based scoring, though server-side heterogeneity confounds rankings. The mid-range score reflects transparent methodology held back by synthetic-only data and infrastructure-dependent results.
Research summary
Ankit Raj (ORCID 0009-0007-6514-1989) is an early-career researcher whose catalogued work to date consists of two 2026 arXiv preprints spanning natural language processing and AI agent safety with document forensics. The first, 'Open-Jev Judgments on CallScreenBench,' applies a JevLite typed-decision readout to Qwen3-4B for one-pass scam-call screening, achieving judgments non-inferior to an LLM judge at roughly 30x lower latency; Evolve's review notes candid treatment of test-set exposure but flags that all evaluation data is synthetic. The second, 'Agents That Edit Documents,' benchmarks how reliably LLM agents can forge archived PDFs against a non-agentic control, reporting an 81.1% raw success rate that falls to 46.2% under strict filtering; rule-based scoring is judged rigorous, though server-side heterogeneity complicates rankings. On both papers Raj holds a middle authorship position without a corresponding-author role, working within a pool of 10 distinct co-authors including Simiao Ren, Tommy Duong, and Yuxin Zhang. Radar scores place collaboration at 48 and reproducibility at 55 (both papers medium readiness), while output, impact, and activity remain unscored for lack of data.
Collaboration pattern
Both catalogued papers carry a middle authorship position with no corresponding-author role, indicating contribution within larger teams rather than independent leadership. The work involves 10 distinct co-authors, with repeated collaborators including Simiao Ren, Tommy Duong, Yuxin Zhang, Dennis Ng, Xingyu Shen, Kidus Zewde, Yuchen Zhou, and Neo Tiangratanakul. The collaboration score of 48 sits mid-range, consistent with steady team participation across the two 2026 preprints.
Research keywords
Highlights
Calibrated scam-call screening at ~30x lower latency
Measuring agentic PDF forgery against a non-agentic control
Timeline
- 2026Two arXiv preprints enter the Evolve catalog, marking this researcher's first catalogued output.
- 2026Co-authors the NLP preprint on calibrated one-pass scam-call screening with a small language model (Qwen3-4B).
- 2026Co-authors the AI agent safety preprint benchmarking agentic PDF forgery against a non-agentic control.
- 2026Accumulates a collaboration footprint of 10 distinct co-authors across the two 2026 preprints.
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 scores are computed from public bibliographic data (OpenAlex-sourced snapshot); citation, venue, and productivity metrics are absent, leaving output, impact, and activity unscored.
- Institution and country are not resolved for this researcher; affiliation context is unavailable.
- Both catalogued papers are 2026 arXiv preprints; peer-review status is not captured in the snapshot.
- 'Ankit Raj' is a relatively common name; this profile is name-resolved from public bibliographic data via ORCID 0009-0007-6514-1989.