brycewang-stanford PLDI Related Work
At PLDI, related work is an adversarial exercise: your reviewers are frequently the authors of your closest baselines. The section must therefore do two things — prove you know the landscape's mechanisms (not just its citations), and state, line by line, what this paper does that each prior technique cannot.
The delta discipline
Ban survey sentences ("Much work has studied pointer analysis [3,7,12,19]"). Every citation earns its place with a mechanism and a delta:
Steensgaard-style unification [7] scales linearly but merges all callees' effects; our signatures keep callee effects separate at equal asymptotic cost (§4.3).
If you cannot write the delta clause, you have not read the paper closely enough to cite it at PLDI — where the cited author may be reviewer B.
Coverage map
| Neighbor | What to check there | Typical relationship to a PLDI paper |
|---|---|---|
| PLDI, last ~4 cycles | The direct competition; reviewers expect these compared, not just cited | Baseline or head-to-head |
| POPL | The semantic foundations your system builds on | "We implement/scale what [x] formalized" |
| OOPSLA | Language design, empirical SE-flavored PL studies | Alternative evidence styles for similar claims |
| ICFP | Functional-language techniques, typed IRs | Technique transfer across paradigms |
| ASPLOS / CGO / PPoPP | Architecture-facing compilation and parallelism | Same optimization, different evidence bar |
| CAV / TACAS | Verification engines and solvers | Your backend or your competitor's |
| ICSE / FSE | Testing and analysis at software-engineering scale | Scale claims to compare honestly |
Since POPL, OOPSLA, ICFP, and PLDI all publish in PACMPL, "recent PACMPL papers" is a single searchable stream — sweep it by issue name so a POPL 2025 article does not get labeled PLDI 2025 in your bibliography.
Cite the PACMPL era correctly
PLDI 2023 and later papers are journal articles, not proceedings entries:
@article{gopinathan2023proofrepair,
author = {Gopinathan, Kiran and Keoliya, Mayank and Sergey, Ilya},
title = {Mostly Automated Proof Repair for Verified Libraries},
journal = {Proc. ACM Program. Lang.},
volume = {7},
number = {PLDI},
articleno = {107},
year = {2023},
doi = {10.1145/3591221}
}
Pre-2023 PLDI papers keep @inproceedings form with page ranges. Mixing the two
conventions — or citing a PACMPL article as "In PLDI '23, pp. ..." — is the most
common bibliography error since the switch, and PL reviewers notice.
Verification before submission
- Confirm every venue attribution on dblp; the classics are the traps (LLVM is
CGO 2004, CompCert's certification paper is POPL 2006, RustBelt is PACMPL
Issue POPL — see
resources/exemplars/library.mdfor the curated list). - For arXiv-only concurrent work, cite it as a preprint and position gently; it is context, not an archival obligation.
- Double-blind interacts here: your own prior papers appear in third person, cited normally — omitting them is itself an identity signal.
Output format
[Delta audit] citations without a delta clause: <list>
[Coverage] PLDI recent cycles / POPL / OOPSLA / ICFP / systems neighbors — gaps?
[Head-to-head] which cited techniques appear in the evaluation as baselines?
[Bibliography form] PACMPL-era entries in journal form? venues dblp-verified?
[Anonymity] self-citations third person, none omitted?