Horace He

Important for the bridge between early open-model scaling work and later frontier closed-model systems, especially around architecture and training-stack choices that ended up mattering at both ends of the field.

Useful because his footprint runs through the early EleutherAI training stack, GPT-NeoX, and Pythia, which makes the page a better map of open-model infrastructure than a generic one-paper profile.

A useful person to follow for the evaluation layer of open models, especially where benchmark infrastructure and RLHF tooling become reusable community assets rather than one-off lab code.

Worth tracking if you care about the seam between open-model benchmarking and the harder question of what frontier systems should actually be evaluated for.

One of the quieter but still important contributors in the open-data and open-evaluation lineage behind The Pile, GPT-NeoX, and later benchmarking infrastructure.

Important if you care about the European sovereign-AI track, especially the attempt to build multilingual, explainable, and compliance-conscious frontier systems outside the US lab stack.