Diyi Yang

Stanford University · United States

Unclaimed · AI-structured profileORCID OpenAlex

Stanford researcher bridging NLP, human-AI interaction, and the future of work — from topic models to LLM agents

papers here

reviewed by Evolve

first-author

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—

The output axis measures publication volume and productivity trajectories, but this snapshot lacks the underlying per-year output records needed to score it, so the dimension is left unscored rather than estimated from the three reviewed 2026 papers.

Impact—

Impact is not yet scored: the snapshot carries an h-index of 49 and an i10-index of 144, but the citation distributions and field-normalized measures this axis requires are unavailable, so no impact estimate is offered here.

Focus0/100

Focus scores 0 because indexed work splits evenly across three fields — machine learning (human-AI interaction and LLM agents), HCI/human-AI interaction, and human-AI collaboration and the future of work — at 0.333 each. Topic counts confirm the same breadth: Topic Modeling 114, Natural Language Processing Techniques 81, Speech and Dialogue Systems 43, Sentiment Analysis and Opinion Mining 33, and Ethics and Social Impacts of AI 33.

Collaboration80/100

Scores 80: the indexed record lists 53 distinct coauthors alongside 94 works in the last three years, indicating a strongly team-based style. Recurring partners include Zora Zhiruo Wang, Rulin Shao, Apurva Gandhi, Aspen Chen, Qianou Ma, Luxi He, and Michael Ryan, within a frequent-coauthor pool spanning both NLP and HCI communities.

Activity100/100

Scores a perfect 100 for recent activity: 94 works in the last three years, with OpenAlex yearly counts rising from 13 (2020) to 37 (2021), 38 (2022), a peak of 54 (2023), then 44 (2024) and 42 (2025), plus 8 already indexed in 2026 — a steep and continuing productivity climb.

Reproducibility43/100

Scores 43 (reproMean across 3 assessed papers). In the Evolve catalog, readiness is medium for the LLM-as-oracles study and the TAHI test-time adaptation paper, but low for the workflow-level augmentation framework, an empirically untested whitepaper; Evolve also flags LLM-judge circularity and numeric mismatches as concrete reproducibility risks.

Research summary

Diyi Yang is a researcher at Stanford University working at the intersection of natural language processing, human-computer interaction, and the future of work. The OpenAlex-derived snapshot shows an h-index of 49 and an i10-index of 144, with 94 works in the last three years and 53 distinct coauthors. Indexed topics are led by Topic Modeling (114), Natural Language Processing Techniques (81), Speech and Dialogue Systems (43), Sentiment Analysis and Opinion Mining (33), and Ethics and Social Impacts of AI (33). Field assignment splits evenly at 0.333 across machine learning (human-AI interaction and LLM agents), HCI/human-AI interaction, and human-AI collaboration and the future of work — consistent with a radar focus score of 0, reflecting breadth over a single specialization. Activity scores a perfect 100, with yearly output climbing from 13 works in 2020 to a peak of 54 in 2023, 44 in 2024, and 42 in 2025. The three Evolve-reviewed 2026 papers — LLMs as Oracles, a workflow-level framework for AI augmentation, and TAHI test-time adaptation — position this recent work around how people should rely on, personalize, and share work with LLM agents. Reproducibility scores 43, reflecting medium readiness for the empirical studies but low for the empirically untested framework whitepaper.

Collaboration pattern

The record lists 53 distinct coauthors and 94 works over the last three years, backing the collaboration score of 80. The most frequent partners in this window are Zora Zhiruo Wang, Apurva Gandhi, Rulin Shao, Aspen Chen, Jonas Mueller, Zhiqi Liang, Jett Chen, Michael Ryan, Qianou Ma, Luxi He, Zhoujun Cheng, and Andre He — a blend of NLP and HCI collaborators. All three Evolve-cataloged 2026 papers are coauthored, with Yang in middle or last author position and none flagged as corresponding author.

Research keywords

Human-AI interactionLLM agentsHuman-AI collaborationTest-time adaptationAI overrelianceLLM-as-judgeTopic modelingSentiment analysis and opinion miningEthics and social impacts of AIFuture of workAI augmentationPersonalization

Highlights

LLMs as oracles: rising reliance for subjective personal questions
A 2026 arXiv study (middle author) argues with multi-dataset, preregistered evidence that reliance on LLMs for subjective personal questions is rising and largely unnoticed, with implications for user autonomy; Evolve finds the evidence compelling but notes it rests on LLM judges and self-selected donors.
TAHI: adapting LLM agents to individuals through interaction
TAHI ('Efficient Test-Time Adaptation through Human-AI Interaction', 2026, middle author) adapts agents to individual users via 20 interaction sessions, reporting 4.5–20.9% gains; Evolve credits a strong human study but flags grader circularity and numeric mismatches that need fixes.
A workflow-level framework for when AI truly augments work
A 2026 last-author workshop whitepaper, 'When Does AI Augment Work?', defines six conditions for genuine AI augmentation through a workflow-level lens, engaging deskilling and meaningful human control; Evolve calls it conceptually solid but empirically untested.

Timeline

2026Published 'LLMs as Oracles: Reliance on LLMs for Subjective Personal Questions' on arXiv — a preregistered, multi-dataset study of rising reliance on LLMs for subjective personal questions (middle author).
2026Released TAHI ('Efficient Test-Time Adaptation through Human-AI Interaction'), adapting agents to individuals via 20 interaction sessions with 4.5–20.9% gains (middle author).
2026Authored 'When Does AI Augment Work?', a last-author workshop whitepaper defining six conditions for genuine AI augmentation at the workflow level.
2026All three cataloged works landed in the same year, spread evenly across the researcher's three field buckets (0.333 each), within a partial-year OpenAlex count of 8 works for 2026.

Papers in the Evolve catalog

Frequent co-authors

Zora Zhiruo Wang×1 Apurva Gandhi×1 Rulin Shao×1 Aspen Chen×1 Jonas Mueller×1 Zhiqi Liang×1 Jett Chen×1 Michael Ryan×1 Qianou Ma×1 Luxi He×1 Zhoujun Cheng×1 Andre He×1

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.

Bibliometrics (h-index 49, i10-index 144, yearly counts, topics, coauthor counts) come from an OpenAlex-derived snapshot only; no other citation databases were consulted.
Only 3 papers carry Evolve review verdicts; judgments on quality and reproducibility rest on this small reviewed subset.
The output and impact radar axes lack the underlying data and are intentionally left unscored, not estimated.
2026 output is partial-year (8 works indexed so far), and this profile is name-resolved from public bibliographic data under ORCID 0000-0003-1220-3983.