AtlasKV

HKUST KnowCompKnowledge graphs & graph computingListed

About this tool

The official implementation of an ICLR 2026 paper encoding billion-scale knowledge-graph triples as key-value vectors for model knowledge augmentation within 20GB of GPU memory. Includes KG2KV data preparation, training, memory and accuracy evaluation scripts, and checkpoints trained on ATLAS-Wiki and ATLAS-CC.

  • Feature 1: KG2KV converts graph triples into QKV training data.
  • Feature 2: Support for billion-scale ATLAS-Wiki and ATLAS-CC graphs and synthetic data.
  • Feature 3: Choose text-embedding-ada-002, text-embedding-3-large or all-MiniLM-L6-v2 encoders.
  • Feature 4: Scripts evaluate memory use, knowledge accuracy and generation relevance.
  • Feature 5: Checkpoints trained for 3K steps are available.

Resource information

Maintainer
HKUST-KnowComp, Associate Professor Yangqiu Song's group.
Type
Skill
Category
Knowledge graphs & graph computing
Runs on
Agent Skills
Who it's for
Researchers studying model knowledge augmentation, graph–model integration and lower-memory methods, including single-GPU reproduction and comparisons.
License at time of listing
MIT
Source
GitHub

TITU Work Market1

Available in the TITU Work Market. Search for the names below in the app to install.

hkust-atlaskv-kg-memory

Publisher
TITU Work
Version
1.0.0
License
MIT
Last updated
10/28/2026
Installs
0