Resultado de búsqueda
1 de ene. de 2023 · Second Thoughts are Best: Learning to Re-Align With Human Values from Text Edits. Ruibo Liu, Chenyan Jia, Ge Zhang, Ziyu Zhuang, Tony X Liu, Soroush Vosoughi. We present Second Thought, a new learning paradigm that enables language models (LMs) to re-align with human values.
- arXiv:2301.00355 [cs.CL]
Second Thoughts Are Best: or, a Further Improvement of a Late Scheme to Prevent Street Robberies is a 1729 pamphlet by Daniel Defoe. He wrote it under the name of Andrew Moreton Esq. , [1] presented as a dissatisfied middle-class old man who was extremely concerned about the increase in criminality around the 1720s.
Specifically, we include the unaligned source as part of the new “context”, and then train an LM to learn how to make sequential edits on the source to produce the target (Figure 2 (b)). This way the model learns how to recover from a value-unaligned, poisoned context during the generation phase.
We present Second Thoughts, a new learning paradigm that enables language models (LMs) to re-align with human values. By modeling the chain-of-edits between value-unaligned and value-aligned text, with LM fine-tuning and additional refinement through reinforcement learning, Second Thoughts not only achieves superior performance in three value ...
Second Thoughts are Best: Learning to Re-Align With Human Values from Text Edits Ruibo Liu1, Chenyan Jia2, Ge Zhang3,4, Ziyu Zhuang1∗, Tony X. Liu 2, Soroush Vosoughi1 1DartmouthCollege, 2Stanford University, 3Beijing Academy of Artificial Intelligence,4University of Michigan,Ann Arbor 1{ruibo.liu.gr, soroush.vosoughi}@dartmouth.edu Abstract
Second Thoughts are Best: Learning to Re-Align With Human Values from Text Edits Ruibo Liu1, Chenyan Jia2, Ge Zhang3, Ziyu Zhuang1∗, Tony X. Liu 2, Soroush Vosoughi1 1Dartmouth College, 2Stanford University, 3University of Michigan, Ann Arbor 1{ruibo.liu.gr, soroush.vosoughi}@dartmouth.edu Abstract
SECOND THOUGHT definition: 1. If you do something without a second thought, you do it without first considering if you should…. Learn more.
