Contrastive-LM/CLM

Contrastive Language Models A System One Model for Fast and Generalizable Decision-Making 🔥 Contrastive Language Models (CLMs) are a new class of System One model trained with a contrastive learning objective that connects states and actions. This repo serves CLM-8B behind a TypeSafe-compatible API. 30M synthetic hard negatives, and post-trained on 1M agentic trajectories. tasks with up to 9× lower latency. With lightweight fine-tuning it sets a new SOTA as a verifier on agentic coding benchmarks: Terminal-Bench 2.1 (87.6%) and DeepSWE (81.6%). reused independently, which makes training and serving cheap and blazing fast! We invite the community to plug it into their own agents and benchmarks! Requires Python 3.10+, Linux and an NVIDIA GPU. Installs everything, including PyTorch and vLLM. Questions may be Noul / Choice / Score objects or plain wire-format dicts, so a request written for TypeSafe replays as client.systemone(state, questions). systemone is built on one primitive: score…