We have 3 of 13 papers
We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.
Not the right person? Other researchers publish under this name.
Continual adaptation of language models can change their output distribution on prompts learned earlier, while retaining every old prompt-answer pair may be undesirable or impossible. We study condition-anchored generative distillation (CAGD): retain a small set of old prompts, use a frozen previous model to reconstruc...
Results indicate that multi-scale Inception encoding and auxiliary supervision are the most important components, while the PatchX interaction module provides no consistent additional gain under the current setting.
The Personal History Memory Mechanism is introduced, which replaces random perturbation with weighted historical experience to enhance the directional search capability of the algorithm, and a greedy selection strategy is embedded to ensure the monotonically non-deteriorating quality of population solutions.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
