Causal ML researcher focused on abstention-safe, auditable LLM decision support via fuzzy causal evidence fusion
1
papers here
1
reviewed by Evolve
1
first-author
1
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) 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 is unscored: the snapshot provides no publication counts, venue list, or annual volume for this researcher. This axis would measure cumulative scholarly production — number of papers, venues, and the balance between arXiv preprints and peer-reviewed work over time.
Impact—
Impact cannot be scored: no citation counts, h-index, or reference uptake are present in the input. This axis would measure citation-based influence — citations per paper, aggregate citations, and adoption of the F-DACE-style causal-fusion approach by subsequent work.
Focus100/100
Focus scores 100 because the field distribution is fully concentrated: 1 of 1 cataloged papers falls in a single field, machine learning — causal inference and decision support. The F-DACE paper spans fuzzy evidence fusion, double machine learning, heterogeneous treatment effects, and abstention-aware selective prediction — a tightly defined agenda, though the base is a single paper.
Collaboration0/100
The collaboration axis scores 0 from raw data showing 0 distinct co-authors and an empty frequent-coauthor table. The single catalog paper is first-authored and corresponding, so Dey leads the work, but no co-author names are available to characterize network reach; the score reflects a data gap as much as a working style.
Activity—
Activity is unscored: the snapshot lacks per-year publication records beyond a single 2026 entry. This axis would measure recent research tempo — outputs in the last three years, cadence of preprints, and continuity of first-author contributions.
Reproducibility55/100
Reproducibility scores 55, the mean readiness rating across exactly 1 assessed paper (reproMean 55, reproPapers 1). F-DACE carries a medium repro-readiness label in the catalog; its abstention-safe, auditable design supports re-execution, but with one evaluated paper this axis is a preliminary signal rather than a stable track record.
Research summary
Sourish Dey appears in Evolve's catalog as an early-stage profile anchored by a single, sharply defined piece of work. The 2026 arXiv preprint "F-DACE: Fuzzy Disagreement-Aware Causal Evidence Fusion for Abstention-Safe Conversational Retail Decision Support" is first-authored and corresponding-authored, placing Dey at the center of the contribution. The paper sits in machine learning — causal inference and decision support and assembles fuzzy evidence fusion, double machine learning, heterogeneous treatment effects, and causal discovery refutation tests into a pipeline that lets a conversational agent give retail advice while abstaining whenever the underlying causal evidence disagrees. Evolve's review highlights a fail-closed LLM agent design and observes that the framework's honest results tie a unanimity baseline, adding auditability over accuracy. Consistent with this single-paper base, the focus dimension scores 100 (one field, one paper) and reproducibility scores 55, a medium readiness rating averaged over one assessed paper. The output, impact, and activity dimensions remain unscored because the snapshot carries no publication-volume, citation, or per-year productivity data; no affiliation, country, or ORCID is attached to the profile, and the co-author fields are empty. This profile is name-resolved from public bibliographic data.
Collaboration pattern
The snapshot lists 0 distinct co-authors and an empty frequent-coauthor table, so no collaboration network can be described yet; the collaboration dimension accordingly scores 0. The single catalog paper is first-authored and corresponding, indicating Dey leads the work rather than sitting in a middle-author position. Whether this reflects a genuinely solo working style or simply missing co-author metadata cannot be determined from the current data.
Research keywords
Highlights
F-DACE: abstention-safe conversational retail decision support
Fail-closed LLM agent design
Honest results tie the unanimity baseline
Causal-inference method stack with medium repro-readiness
Timeline
- 2026Publishes F-DACE on arXiv as first and corresponding author — fuzzy disagreement-aware causal evidence fusion for abstention-safe conversational retail decision support.
- 2026Evolve reviews the paper, noting a fail-closed LLM agent and honest results that tie the unanimity baseline, adding auditability over accuracy.
- 2026The paper enters the Evolve catalog with medium repro-readiness, anchoring an early-stage, single-field profile (focus 100, reproducibility 55).
Papers in the Evolve catalog
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.
- All scores rest on a single-paper snapshot: focus 100 and reproducibility 55 are computed from one 2026 arXiv paper in the reviewed catalog.
- Output, impact, and activity dimensions are unscored because publication counts, citation data, and per-year records are absent from the input.
- No institutional affiliation, country, or ORCID is attached; this profile is name-resolved from public bibliographic data.
- Co-author fields are empty (0 distinct co-authors), so the collaboration dimension reflects a data gap rather than a verified solo pattern.