Research

Aveo Research Labs is a research-first organization. Below is a selection of peer-reviewed work authored by our founders and team.

Founding research footprint
29,500+citations
50+publications
2Test of Time Awards
1Nature Mach. Intell. first-author

HETEROGENEOUS GRAPH LEARNING

  • Heterogeneous Graph Transformer — WWW 2020
  • Pre-training Graph Neural Networks for Generic Structural Feature Extraction — ICLR 2019

GRAPH NEURAL NETWORKS & MINING

  • Meta-path based heterogeneous information network mining — KDD (Test of Time Award)

NETWORK SCIENCE

  • FINDER: Finding key players in complex networks through deep reinforcement learning — Nature Machine Intelligence
Influences & Benchmarks — Not Our Work

Our methodology builds on and is calibrated against the following external work. We acknowledge these as the shoulders we stand on — they are not our publications.

Foundational methodology

Representation EngineeringZou et al., arXiv:2310.01405 — foundational steering vector framework
Activation SteeringRimsky et al., arXiv:2308.10248 — practical activation intervention
SimCLRA Simple Framework for Contrastive Learning of Visual Representations — ICML 2020
AutoGenMulti-Agent Conversation Framework — 2023

Validation benchmarks

OmniBehaviorarXiv:2604.08362 — prompt-only simulation accuracy ceiling
SCOPEarXiv:2601.07110 — demographics explain 1.5% of behavioral variance

Causal inference foundations

Unit Selection ProblemPearl & Li, arXiv:2210.08203 — counterfactual classification framework