GRACE dynamically constructs and updates coresets for LLM training using representation diversity, gradient-based importance, and k-NN graph propagation to improve efficiency and performance.
Keyon Vafa, Justin Y Chen, Ashesh Rambachan, Jon Kleinberg, and Sendhil Mullainathan
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 2roles
background 1polarities
background 1representative citing papers
NextLat augments next-token prediction with latent next-state prediction, theoretically converging latents to belief states and showing empirical gains in world modeling, reasoning, planning, and faster inference via speculative decoding.
citing papers explorer
-
GRACE: A Dynamic Coreset Selection Framework for Large Language Model Optimization
GRACE dynamically constructs and updates coresets for LLM training using representation diversity, gradient-based importance, and k-NN graph propagation to improve efficiency and performance.
-
Next-Latent Prediction Transformers Learn Compact World Models
NextLat augments next-token prediction with latent next-state prediction, theoretically converging latents to belief states and showing empirical gains in world modeling, reasoning, planning, and faster inference via speculative decoding.