Experiments across code LLMs show no-review collapses fastest, human-gated filters slow collapse, and AI self-gates lose effect over time, degenerating to ungated self-training under self-confirming acceptance as proven via gated distributional reweighting and spectral analysis.
arXiv preprint arXiv:2212.14402 , year=
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
representative citing papers
Hidden activations in LLMs encode detectable information about statement truthfulness, enabling a classifier to identify true versus false content more reliably than the model's assigned probabilities.
A hackathon study with no-manual-edit rules found that vibe coding engagement varies by skill level and task complexity, with implications for using AI in programming education.
A survey of nearly 1000 NLP & Law papers from 2013-2024 documenting increases in publication volume, scope, methodological sophistication, and data/code availability.
citing papers explorer
-
When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs
Experiments across code LLMs show no-review collapses fastest, human-gated filters slow collapse, and AI self-gates lose effect over time, degenerating to ungated self-training under self-confirming acceptance as proven via gated distributional reweighting and spectral analysis.
-
The Internal State of an LLM Knows When It's Lying
Hidden activations in LLMs encode detectable information about statement truthfulness, enabling a classifier to identify true versus false content more reliably than the model's assigned probabilities.
-
Code for All: Educational Applications of the "Vibe Coding" Hackathon in Programming Education across All Skill Levels
A hackathon study with no-manual-edit rules found that vibe coding engagement varies by skill level and task complexity, with implications for using AI in programming education.
-
Natural Language Processing in the Legal Domain
A survey of nearly 1000 NLP & Law papers from 2013-2024 documenting increases in publication volume, scope, methodological sophistication, and data/code availability.