Topic: repair

A new method called EditVeto has been proposed to address the issue of language models altering correctly transcribed text when correcting OCR outputs of historical documents. This technique acts as a veto gate after full-context rewriting by LLMs, ensuring only necessary edits are made.

arxiv.org↗

A new method called HyperRepair has been developed to recover the general capabilities of fine-tuned language models without compromising their specialized performance, addressing the common trade-off faced by existing techniques. This breakthrough could enhance the versatility and reliability of language models across various applications.

arxiv.org↗

Intelligent agents should self-improve by learning from their mistakes, but current systems forget lessons quickly, hindering long-term progress. This issue highlights the need for more robust mechanisms to ensure agent experiences lead to lasting improvements.

arxiv.org↗