Hung-Yu Lin

Research profile

Research summary

Hung-Yu Lin is affiliated with Texas A&M University–Corpus Christi (ORCID 0000-0002-9035-5408). This profile is name-resolved from public bibliographic data, so identity should be treated as probable rather than confirmed. The Evolve-reviewed work sits in two fields — AI evaluation methodology and AI for education / educational technology — each carrying 0.5 of the profile's field distribution. In 2026 the researcher published a first-author arXiv position paper arguing that compliance data from deployed automated-driving systems, such as disengagement reports, cannot alone support comparative safety claims; it proposes an 'evaluation contract' for measurement validity, which Evolve judges persuasive but untested, with low reproducibility readiness. A second 2026 preprint, LLMersion, describes a local-first agent framework built on small open models with on-device inference for low-cost home language learning toward educational equity; Evolve finds the feasibility argument strong and efficacy untested, with medium reproducibility readiness. The bibliometric snapshot records an h-index of 30, an i10-index of 73, 25 works in the most recent three years and an activity score of 95/100, but its top OpenAlex topic tags are biomedical and conflict with the AI-oriented field distribution — an inconsistency we flag as likely name-resolution noise. Output and impact dimension scores are missing in this snapshot and are left unscored.

Research keywords

Highlights

• Compliance data is not a substitute for evaluation

  In a 2026 first-author arXiv paper, Lin argues that compliance artifacts from deployed automated-driving systems — such as disengagement reports — cannot on their own support comparative safety claims. Evolve calls the paper a persuasive position piece whose proposed 'evaluation contract' is coherent but untested, and rates its reproducibility readiness low.

• LLMersion: a local-first AI tutor for language-learning equity

  The 2026 LLMersion preprint proposes a free, local-first agent framework built on small open models with on-device inference, supporting all four language skills for home learners. Evolve finds the educational-equity and feasibility argument well argued while efficacy remains untested, and rates reproducibility readiness medium.

• An 'evaluation contract' for deployed AI systems

  The same 2026 position paper frames measurement validity for deployed AI through an 'evaluation contract' that connects what compliance data records to what comparative safety claims actually require. Evolve deems the framing coherent and persuasive while noting it remains empirically untested — a concrete agenda for future benchmark work.

Timeline

1. 2003Earliest year appearing in this profile's OpenAlex countsByYear, with 1 recorded work.
2. 2016countsByYear records 6 works in 2016, beginning a sustained climb through 9 works in 2017.
3. 201916 works recorded in 2019, the highest annual total up to that point in the OpenAlex counts.
4. 2021Peak productivity year: 22 works recorded in 2021, the highest single-year count in the profile's countsByYear.
5. 2026Two arXiv preprints enter the Evolve catalog: a first-author measurement-validity position paper for deployed AI, and LLMersion, a local-first framework for low-cost home language learning.

Papers in the Evolve catalog

• 2026-09-12first author

  When Compliance Data Masquerades as Evaluation: Measurement Validity for Deployed AI Systems

• 2026-09-12

  LLMersion: A Local-First AI Agent Framework for Low-Cost Home Language Learning toward Educational Equity

Frequent co-authors