Speech AI researcher working on audio reasoning benchmarks and fairness auditing of speech-to-speech models.
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
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 volume cannot be scored yet: the Evolve catalog holds only two 2026 arXiv papers for this name, with no publication counts or venue history, so productivity is not measurable from the current snapshot.
Impact—
Impact is unscored: no citation counts, h-index, or downstream-usage data are available for these two recent 2026 preprints, so this axis would measure citations and adoption once the bibliometric record accumulates.
Focus0/100
The focus score is 0 because the field distribution splits exactly evenly: 50% speech and natural language processing (audio-language models) and 50% speech processing / AI fairness. The two 2026 papers sit one in each area, so the record shows deliberate two-track work — evaluation benchmarks on one side, fairness auditing of speech-to-speech systems on the other — rather than concentration in a single niche; a larger corpus would show whether this breadth consolidates.
Collaboration41/100
The collaboration score is 41, drawn from 7 distinct coauthors across the two catalog papers. The frequent-coauthor list — Abhishek Mukherji (appearing on both entries), Asif Shaik, Rachuri Lokesh, Sushanta Kumar Pani, Xiaoqun Liu, and Tanu Mitra — suggests small, repeat teams rather than large multi-institution consortia. Neither paper lists this researcher as corresponding author, and authorship ranges from first (SEABED) to middle (the fairness audit).
Activity—
Recent activity is not yet scoreable: both known papers are dated 2026 and no per-year output series exists, so this axis, which tracks publication cadence over time, currently lacks the data it needs.
Reproducibility55/100
Reproducibility averages 55 across the 2 scored papers, with both rated 'medium' reproducibility readiness. The SEABED benchmark ships a reusable evaluation resource, though LLM-authored gold labels without human validation limit trust in the answers; the fairness audit reports tests of five speech-to-speech models but notes no code or data release and only one voice per gender, which constrains independent replication.
Research summary
Harshit Rajgarhia is an early-stage researcher in speech and natural language processing whose Evolve catalog centers on two 2026 arXiv papers at the intersection of audio-language model evaluation and speech-system fairness. The first-authored SEABED (SouthEast Asian Benchmark for Evaluating Audio Reasoning) introduces a reasoning-first benchmark for audio-language models across Southeast Asian languages, touching low-resource speech, prosody, and dialect identification; Evolve's review finds the best model reaches only 50.3%, and credits rigorous ablations while flagging LLM-authored gold labels and the absence of human validation. The second paper, 'Voice or Stereotype? Disentangling Acoustic and Content-Based Gender in Speech-to-Speech Models,' is a fairness audit arguing that bias in speech-to-speech systems lies in gender attribution rather than voice rendering, with five tested models misgendering up to 90% of cases; the review finds the argument convincing but notes a single voice per gender and no code or data release. Field distribution splits evenly (50%/50%) between the two areas, yielding a focus score of 0, and work so far involves 7 distinct coauthors — a collaboration score of 41 — with Abhishek Mukherji recurring among frequent coauthors. Reproducibility averages 55 over the two scored papers, both rated medium. No affiliation, country, or ORCID is on record, and output, impact, and activity axes await more data.
Collaboration pattern
The two catalog papers share a small coauthor pool: 7 distinct coauthors total, with Abhishek Mukherji recurring in the frequent-coauthor list alongside Asif Shaik, Rachuri Lokesh, Sushanta Kumar Pani, Xiaoqun Liu, and Tanu Mitra. This points to compact, repeat collaborations rather than large rotating consortia. Authorship spans first author on SEABED and a middle position on the fairness audit, with no corresponding-author role recorded; the collaboration score is 41.
Research keywords
Highlights
SEABED: a reasoning-first audio benchmark for Southeast Asian languages
Reasoning-first evaluation design backed by rigorous ablations
Bias in speech-to-speech models hides in gender attribution
Timeline
- 2026Releases SEABED as first author on arXiv — a SouthEast Asian benchmark for evaluating audio reasoning in audio-language models.
- 2026Co-authors the fairness audit 'Voice or Stereotype?', showing five speech-to-speech models misgender up to 90% of cases.
- 2026Both catalog papers appear in the same year, splitting evenly between audio-language model evaluation and speech-processing fairness.
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 come from the Evolve bibliometric snapshot (OpenAlex-based public data); both catalog papers are arXiv preprints, so peer-reviewed venue status is not yet reflected.
- No institutional affiliation, country, or ORCID is on record; the profile is name-resolved from public bibliographic data.
- Output, impact, and activity are unscored (missing-data axes) because only two 2026 papers and no citation history are available.
- The timeline covers 2026 only; no earlier publications or career history appear in the provided data.