Bojun Lin

—

Unclaimed · AI-structured profileOpenAlex

Emerging researcher in LLM agent memory and statistically honest evaluation; last author on 2026 arXiv preprints

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-27) 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—

This axis would measure total publication volume over time. The underlying data is insufficient to score it: OpenAlex indexes only one work in the last three years with an h-index of 0, so no output estimate is offered here.

Impact—

This axis would measure citation-based influence (h-index, i10-index, citation trajectories). Available data is insufficient: both indices are 0 with no citation history, so no impact score is provided rather than guessing from thin signals.

Focus—

This axis would measure thematic concentration across an author's works. It cannot be scored: OpenAlex lists three topics with one work each, and they do not align with the two reviewed papers' themes, leaving the research focus unresolved.

Collaboration44/100

Scoring 44/100, collaboration is the stronger of the two measurable axes. OpenAlex records 8 distinct coauthors, and the frequent-collaborator list includes Yezhou Cheng, Runjia Du, Zeming Liu, Qibai Chen, Hang Lyu, Yankai Zeng, and Yilan Wei (listed twice). Both reviewed 2026 papers are multi-authored with Lin in the last position, a pattern consistent with group projects, though the moderate score reflects a small overall network so far.

Activity20/100

Scoring 20/100, recent activity is low in the indexed record: recent3yWorks is 1 and countsByYear contains a single year, 2026, with one work. Both reviewed papers carry a 2026 date on arXiv, suggesting publishing activity that bibliographic indexing has not yet fully captured; the score should be read as an early-record snapshot rather than a long-term trend.

Reproducibility—

This axis would measure author-level reproducibility signals such as code and data availability. No underlying data exists yet; Evolve's per-paper readiness ratings (high, medium) are paper-level judgments, not an author-level score, so the axis stays unscored.

Research summary

Bojun Lin is an emerging researcher whose verifiable public record centers on two 2026 arXiv preprints reviewed by Evolve, both with Lin in the last author position and neither marked as corresponding author. The first, 'Scope Before You Persist', proposes scoping retrieval before writing to persistent agent memory to prevent cross-family interference; Evolve finds the paired evidence solid but limited to a single templated benchmark and one model. The second, 'Sharp Limits for Honest Uncertainty in Hard-Budget Repeated Evaluation', proves sharp replication limits for hard-budget LLM evaluation with matching designs and equal-budget evidence, and is rated high for reproducibility readiness. OpenAlex currently indexes one work in the last three years, with an h-index of 0 and an i10-index of 0, alongside eight distinct coauthors — consistent with a very early or recently started publishing record. Notably, the OpenAlex topic labels on the indexed record — robotics and sensor-based localization, social robot interaction and HRI, and multimodal machine learning — do not match the themes of the reviewed papers, so identity disambiguation remains uncertain and this profile is name-resolved from public bibliographic data. Taken together, the record points to a researcher interested in the reliability of LLM systems — both in what agents remember and in how honestly their evaluation budgets are spent — although the evidence base remains thin.

Collaboration pattern

Lin publishes in teams: both reviewed 2026 papers are multi-authored, with Lin in the last author position and no corresponding-author marking. OpenAlex records 8 distinct coauthors, and the recurring collaborator list includes Yezhou Cheng, Runjia Du, Zeming Liu, Qibai Chen, Hang Lyu, Yankai Zeng, and Yilan Wei — the last appearing twice, likely a duplicate entry. The collaboration score of 44/100 reflects a real but still small network.

Research keywords

LLM agentspersistent memoryself-improvementretrieval scopecross-family interferencecontinual adaptationrepeated evaluationconfidence intervalsminimax lower boundshard budget constraintsLLM benchmarkingstratified sampling

Highlights

Scoped retrieval to stop cross-family interference in agent memory
The 2026 arXiv paper 'Scope Before You Persist' argues that retrieval should be scoped before persisting to memory, and reports paired evidence that this prevents cross-family interference; Evolve judges the evidence solid but notes it comes from one templated benchmark and a single model, with medium reproducibility readiness.
Sharp replication limits for hard-budget LLM evaluation
In 'Sharp Limits for Honest Uncertainty in Hard-Budget Repeated Evaluation' (2026), the team proves minimax lower bounds for repeated evaluation under hard budget constraints, with matching designs and equal-budget evidence; Evolve notes practical gains are regime-dependent and rates reproducibility readiness high.
A common thread: the reliability of LLM systems
Across both reviewed 2026 papers, the shared concern is whether LLM systems can be trusted — one keeps an agent's persistent memory free of cross-family interference, the other bounds what repeated evaluation under hard budgets can honestly conclude — with Lin in the last author position on each.

Timeline

2026Posted the arXiv preprint 'Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory' as last author.
2026Posted the arXiv preprint 'Sharp Limits for Honest Uncertainty in Hard-Budget Repeated Evaluation' as last author; Evolve rates its reproducibility readiness high.
2026OpenAlex indexes one work for this author in this single year, with 8 distinct coauthors on record.

Papers in the Evolve catalog

Frequent co-authors

Yezhou Cheng×2 Runjia Du×2 Zeming Liu×2 Qibai Chen×2 Hang Lyu×2 Yankai Zeng×2 Yilan Wei×1 Yilan Wei×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.

Bibliometric figures come from OpenAlex only; no institutional affiliation, country, or ORCID is attached to this record.
'Bojun Lin' is name-resolved from public bibliographic data; with an h-index of 0 and one indexed work, identity confidence is limited and the record may be incomplete.
OpenAlex topic labels (robotics and sensor-based localization, social robot interaction and HRI, multimodal machine learning) diverge from the reviewed papers' themes (LLM agent memory; evaluation statistics), suggesting possible name-disambiguation error.
The evidence base is limited to two Evolve-reviewed arXiv preprints from 2026; no grant, award, or career history data was available.