Pinned
a man and a woman are standing next to each other and the man is saying so this is how liberty dies
ALT: a man and a woman are standing next to each other and the man is saying so this is how liberty dies
media.tenor.com
One thing I’ve noticed with newer LLMs is inconsistency in self-critique.
If you ask the LLM to ruthlessly critique content it wrote (which is not known to it), it tears it apart.
But once you say "you generated it", LLM backtracks. That suggests bias toward “self,” not pure quality evaluation.
If you ask the LLM to ruthlessly critique content it wrote (which is not known to it), it tears it apart.
But once you say "you generated it", LLM backtracks. That suggests bias toward “self,” not pure quality evaluation.
Reposted by Nitigya
If you come across generative A.I. images in journals, please let me know using this form:
bit.ly/JournalSlop
#AcademicChatter
bit.ly/JournalSlop
#AcademicChatter
Slop graphics in academic journals
Use this form to share graphics (e.g., figures, covers, graphical abstracts) made using generative A.I. that have appeared in academic journals.
The plan is to compile these slop graphics in an ope...
bit.ly
December 18, 2025 at 5:48 PM
If you come across generative A.I. images in journals, please let me know using this form:
bit.ly/JournalSlop
#AcademicChatter
bit.ly/JournalSlop
#AcademicChatter
One thing I’ve noticed with newer LLMs is inconsistency in self-critique.
If you ask the LLM to ruthlessly critique content it wrote (which is not known to it), it tears it apart.
But once you say "you generated it", LLM backtracks. That suggests bias toward “self,” not pure quality evaluation.
If you ask the LLM to ruthlessly critique content it wrote (which is not known to it), it tears it apart.
But once you say "you generated it", LLM backtracks. That suggests bias toward “self,” not pure quality evaluation.
a man and a woman are standing next to each other and the man is saying so this is how liberty dies
ALT: a man and a woman are standing next to each other and the man is saying so this is how liberty dies
media.tenor.com
December 16, 2025 at 6:54 PM
One thing I’ve noticed with newer LLMs is inconsistency in self-critique.
If you ask the LLM to ruthlessly critique content it wrote (which is not known to it), it tears it apart.
But once you say "you generated it", LLM backtracks. That suggests bias toward “self,” not pure quality evaluation.
If you ask the LLM to ruthlessly critique content it wrote (which is not known to it), it tears it apart.
But once you say "you generated it", LLM backtracks. That suggests bias toward “self,” not pure quality evaluation.