The AI slop tax: are careful authors losing the benefit of the doubt at ICLR?
tughanbulut · reddit · 2026-11-09
A citation-checking site operator distills a concern from ICLR discussion threads: after seeing too much unchecked AI-generated writing, are reviewers losing the benefit of the doubt for careful authors?
Key signals:
- One reviewer admitted suspected AI-generated writing puts them off "even if the findings are legitimate."
- An author reporting 4/2/2/2 scores said one review addressed a completely different paper — a one-sided account, but it raises the question of what actually gets a factual correction in a review properly considered.
The worry: reviewer frustration carries over — does an unclear sentence or citation error start looking like evidence the authors never understood their own work? The author argues that when an author can show an objection rests on something the paper doesn't say, there should be a mechanism to get that specific point reconsidered.
They ask reviewers and ACs whether they've grown more suspicious this cycle and how to distinguish correctable mistakes from deeper flaws, linking to a longer field note collecting the original threads, policies, and rebuttal guidance.
More from AGI Musings
- Same AI, different game: getting answers vs directing engineering work — burhop · 2026-11-09
- Only 18% of US firms use AI, but federal statistics can't yet track its labor impact — robseamans · 2026-11-09
- METR task-horizon debate: logistic rollover by 2027 or hyperbolic singularity? — teortaxesTex · 2026-11-09
- 7-year CV engineer on the genAI hype: where are the hard non-LLM problems? — SchwingBesen · 2026-11-09
- VC argument: in the AI slop fest, entrepreneurs hold the leverage, not the rich — StewartalsopIII · 2026-11-09
- LLMs will scale apocryphal attribution losses across the humanities — StewartalsopIII · 2026-11-09