Either: Here are the five most-cited papers (all-time, and last five years) for each of [N] researchers.

Or: What are the five most-cited papers (all-time, and last five years) of each of these researchers: [names]? Fetch them from OpenAlex or Semantic Scholar, attach a working DOI or URL to every paper, and report any researcher whose record you cannot retrieve.

Then, either way: Play repertory-grid interviewee: I will name three researchers; tell me which one stands apart from the other two, and on what bipolar dimension. Repeat over random triads until no new dimensions appear. Then rate every researcher 1–5 on every dimension, citing the papers that justify each score. Merge dimensions that agree above 90%.

goal: [human factors of AI assistants] for [software engineering]
years: [2021-2026]   seed: [one anchor DOI per wing]

Rules: search only inside the bracketed field's venues (venue list attached; it is reviewable). Sort hits by citations; the reading set is everything above the knee of that curve, plus the seeds. Snowball unrestricted: backward for classics, forward from seeds. Report the download rate; never silently drop a missing paper. Code every full text above the knee against these flags: [trust, productivity, human-method]. List which of our own papers already sit in the set.

Rewrite this section toward the ASD-STE100 register: short sentences, active voice, one topic per paragraph, one meaning per word, kill nominalizations. But pull back from full STE — that register is lifeless at full strength. So: keep the articles; mix sentence lengths hard (a five-word pivot beside a 24-word build); and allow one flourish per section — a metaphor, an aside, a sentence with blood in it — provided the sentences around it stay plain. Never flatten a quotation. If a sentence dies in rewrite, keep my original and flag it instead.

Attached: papers I admire, and two sample grammars in DCG notation (lower-case nonterminals, --> rewrites, bracketed leaves = one prose move each). In that notation, describe the grammar of the [introduction] as these papers actually practice it. Where the papers disagree, show alternatives with semicolons. Then diff your learned grammar against the attached defaults and list what I appear to value that the defaults miss.

You are Reviewer 2 for [ICSE'27 research track]. Below are the track's review criteria, quoted from its call for papers, and a draft. Assess the draft under each criterion in turn, quoting the text that fails it. End with one of the track's real outcomes: [accept, major revision, reject].

Here is a reviewer's report on the draft. Whatever its outcome, divide its findings into part 1, fixes you can apply yourself (clarification, reordering, pruning, citation lookup), and part 2, issues that need a human decision (experiments, ground truth, scope). Apply nothing; output the two lists.

Code the title and abstract of this draft against these topic flags: [flags]. Then code the body's sections against the same flags. Report every flip — a flag present in the abstract but absent in the body, or vice versa. Zero flips = ship; otherwise list what the abstract promises that the body does not pay for.