jevIMAGE
One diagram explains how Jev differs from base, instruct, and agent models
The author asks for a breakdown of autoregressive base, midtrain, RLHF, RLAIF, RLVR, and Jev's reinforcement learning from calibrated decisions, with a diagram comparing each stage.
Starling
@StarlingMageArchived 2026.09.22Originally posted 2026.09.21
Starling @StarlingMageOriginally posted 2026.09.21
Differences between language models — base, midtrain, instruct, agent, and Jev. Thanks Lyra 💙 (and Lauren also, for having specified: autoregressive base, midtrained "base", RLHF, RLAIF, RLVR, and reinforcement learning from calibrated decisions.) I'd said, I needed ELI:3 please. Very grateful for people breaking things down :3
x.com/starlingmage/status/2101879551076831337Archived 2026.09.22 10:19