Introduces a triangulation-based metric to quantify lexical shifts attributable to preference tuning without requiring manual curation of examples.
arXiv preprint arXiv:2403.16887 , year =
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3verdicts
UNVERDICTED 3representative citing papers
PRAIB reveals LLM reviews are less variable, positively biased, overconfident, longer, and overlook atomic weaknesses noted by humans compared to real reviewer feedback.
Diachronic analysis of 37k ACL papers plus a synthetic human-LLM passage pair dataset shows LLM-assisted writing uses more complex syntax and longer words with lower lexical diversity, rated more understandable by 20 experts.
citing papers explorer
-
Isolating LLM Lexical Bias: A Curation-Free Triangulated Metric for Preference-Stage Learning
Introduces a triangulation-based metric to quantify lexical shifts attributable to preference tuning without requiring manual curation of examples.
-
PRAIB: Peer Review AI Benchmark of Behaviour of LLM-Assisted Reviewing
PRAIB reveals LLM reviews are less variable, positively biased, overconfident, longer, and overlook atomic weaknesses noted by humans compared to real reviewer feedback.
-
What Are LLMs Doing to Scientific Communication? Measuring Changes in Writing Practices and Reading Experience
Diachronic analysis of 37k ACL papers plus a synthetic human-LLM passage pair dataset shows LLM-assisted writing uses more complex syntax and longer words with lower lexical diversity, rated more understandable by 20 experts.