§ feed · storyline
Contrastive-LM's CLM-8B model scores agent actions 9× faster than Jev
Contrastive-LM releases CLM-8B, an open model that scores candidate agent actions up to 9× faster than Jev in zero-shot tests, running on a single NVIDIA GPU with cached state/action vectors.
Contrastive-LM released CLM-8B, an open model that scores candidate agent actions rather than generating text. MarkTechPost reports up to 9× lower latency than Jev in zero-shot tests, with cached state/action vectors and self-hosting on one NVIDIA GPU.
§ sources1 publication · timeline below