I would rather quit NLP than read another paper like this: The rise of antithesis in NLP papers

Key points

• Two annotators, blind to the source, find almost no 2019 use annoying and about one in ten 2026 uses; they seldom agree on which, yet about half of 2026 papers contain a use that annoys each of them.1
• Annoying uses present the rejected alternative less favorably than legitimate uses.1
• Raters of preference data and open reward models favor the construction, and an instruction to be honest promotes it.1
• We conjecture that it is a side effect of post-training on pairwise preferences, which credit a disavowal in a single response and cannot register its cost across a text.1

Sources (1)

• [1]I would rather quit NLP than read another paper like this: The rise of antithesis in NLP papers
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 01:50 PM

  We study the construction rather than in ACL papers from 2019, ACL-style arXiv papers from 2026, and papers written by GPT models from the same titles and abstracts.

  Two annotators, blind to the source, find almost no 2019 use annoying and about one in ten 2026 uses; they seldom agree on which, yet about half of 2026 papers contain a use that annoys each of them.

Extractive summary: sentences quoted from the sources.