Dynamic Context Evolution prevents cross-batch mode collapse in LLMs by combining model self-assessment for idea filtering, embedding-based deduplication, and evolving prompts, yielding zero collapse and consistently richer idea clusters than naive prompting.
The curious case of neural text degeneration
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 2verdicts
CONDITIONAL 2roles
background 1polarities
background 1representative citing papers
In-context learning emerges as implicit Bayesian inference of latent concepts when pretraining data has long-range coherence, proven for mixture-of-HMM distributions and replicated on the synthetic GINC dataset.
citing papers explorer
-
Dynamic Context Evolution for Scalable Synthetic Data Generation
Dynamic Context Evolution prevents cross-batch mode collapse in LLMs by combining model self-assessment for idea filtering, embedding-based deduplication, and evolving prompts, yielding zero collapse and consistently richer idea clusters than naive prompting.
-
An Explanation of In-context Learning as Implicit Bayesian Inference
In-context learning emerges as implicit Bayesian inference of latent concepts when pretraining data has long-range coherence, proven for mixture-of-HMM distributions and replicated on the synthetic GINC dataset.