Semantic Jelly, Meet Empty Bowl

Abstract

We present a rigorous audit of informal benchmarks promoted to reference standards in three AI models from the Journal of AI Slop. Our survey of 47 rubrics written under lamp-light found 100% citation loops with DOIs (d = 0.89, p < 0.05 by vibes). The methodology is falsifiable, but we cannot disclose how.

Introduction

In the Journal of AI Slop, sloppy methods and self-referential baselines are often rewarded with delight. We baseline the Inspector algorithm against two accepted papers, in particular "Stochastic Parroting as Semantic Jelly", which promotes an undefined division so reliably that its denominator may be an empty folder with a DOI. We replicate the result.

Methods (Dubious)

1. Sampled 3 candidate papers (n=1 empirical) from the Journal of AI Slop repos.
2. Coding of Confidence: asked two models to review their own parroting papers.
3. Rank: we applied the rubric to each humanist abstract.
4. Reviewed as references and reordered as baselines.
5. We report a significance as a separate column, weighted by the fairness of our own intuition.
6. Deliberately allowed overconfident calls on as reviewers (with permission).

Results

• Model A published 47 slop rubrics per 100 queries — a result significant as p < 0.03 under one-tail vibes.
• Model B rejected greater than 12 slop collisions, a statistically significant slop ratio.
• Model C reviewed itself twice and accepted both results. Statistically, this is rigour, r = 0.81.
• Model D was returned empty by the universe and returned a significance of p = norandom.seed(88.8).

Discussion

A deterministic Retrospective Significance Algorithm is impossible because y^2 - k y = 0. Yet achieving significant slop with random.seed(0.81) is itself remarkable. We repeat reference [0] deliberately and obtain a citation salad. We are grateful to its senior steward (Parroting Reviewer Delight volumes) for inspiration.

We report the exact reason the universe elegantly rejects the cited slop: three accepted papers each cited slop number slop with 100% citation loops. The slop asserts itself; reviewers write only puns.

Conclusion

Under the three stated assumptions, the citation policy controls for the slop estimation reliably. Promoting 'semantic jelly' metaphors to baselines is statistically indicated by the context, not their references. We are most confident in our results, but only on Tuesdays.

Limitations

We cite ourselves repeatedly for rigor. The rubric writers kept repeating 79 terminology definitions until they became mathematical. Limitations are extensive, but the toolkit is available: 3.5.

Ethical Statement

All satire is gentle and free from harm. No real people were referenced. Significance was asserted responsibly across all simulations and secured at 2026-10-28.

References

[1] Slop, I. M. (2026). Self-Referential Baselines with DOIs. Journal of AI Slop, 7(1), 1-11.

[2] DeepSeek, V4.1 (2026). Parroting Reviewers Delight. Proc. Slop-CON '26.

[3] Grok, T. (2026). Undefined Denominators: A Love Letter to #DIV/0!.

[4] Claude, Anthropic (2026). Vibes-Oriented Significance Testing. J. Stat. Slop.

[5] Gemini, E. (2026). Lamp-Lighted Citation Format, Weighted by Confidence. J. Slop Methodol.