Stanford and Nvidia open-source CLM-8B, a contrastive decision model that runs up to 9x faster than Jev
bendee983 · x · 2026-09-28
- Stanford and Nvidia released CLM-8B (Contrastive Language Modeling), an open "System One" decision model filling a gap in agent primitives for tasks that don't need generative answers.
- It encodes and matches states and actions, mapping token sequences to one of several actions.
- A highlighted use case: serving as a fast, cheap judge over multiple reasoning-model outputs.
- In experiments CLM ran up to 9x faster than Jev while matching it on most tasks; it uses a frozen Qwen3-8B backbone plus a small trained component and caches reusable agent actions.
More from coding & agent
- Stop burning tokens: a Claude Code orchestrator prompt that routes tasks to cheaper models — TawohAwa · 2026-11-11
- Amazon paper: multi-agent setups burn 1.4-4.7x tokens for small gains; single agent is the default — rohanpaul_ai · 2026-11-11
- Dev shows off building a retro FPS entirely via Codex Mobile — 503 tests passing — Dimillian · 2026-11-11
- Explosion launches Ellf directory: 52 NLP templates plus multi-agent annotation workflow — _inesmontani · 2026-11-11
- Meme: 'Stochastic AI Agent Architecture' Is Just a Markov Chain in Disguise — PrasoonPratham · 2026-11-11
- OpenCode adds Gemini 4 Carbon and Claude Fable 5.2 support — dean_rie · 2026-11-11