GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior

T2Covered by 1 source · first reported by AI Alignment Forum / LessWrong (curated) at 20 Oct, 20:26 UTC

Authors: Axel Højmark*, Ezra Newman* (*Equal contribution) While reviewing transcripts from standard behavioral evaluations of GLM-5.3, we noticed that the model occasionally reasons, unprompted, about how its actions will shape future models. It appears to consider that its outputs might be used…

ai

Sources

• T2GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future BehaviorAI Alignment Forum / LessWrong (curated)20 Oct, 20:26