H2O evicts non-heavy-hitter tokens from the KV cache using a dynamic submodular policy, retaining recent and frequent-co-occurrence tokens to reduce memory while preserving accuracy.
Gradientcoin: A peer-to-peer decentralized large language models
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
unclear 1representative citing papers
DeRelayL is a proposed sustainable decentralized learning paradigm where permissionless participants relay-train and share models via designed incentives, backed by theoretical analysis and simulations.
citing papers explorer
-
H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models
H2O evicts non-heavy-hitter tokens from the KV cache using a dynamic submodular policy, retaining recent and frequent-co-occurrence tokens to reduce memory while preserving accuracy.
-
DeRelayL: Sustainable Decentralized Relay Learning
DeRelayL is a proposed sustainable decentralized learning paradigm where permissionless participants relay-train and share models via designed incentives, backed by theoretical analysis and simulations.
- Learning Decentralized LLM Collaboration with Multi-Agent Actor Critic