🚀 Introducing Length Value Model (LenVM): a new paradigm for length modeling and a new dimension of scaling. Tokens are the basic unit of inference compute. Length affects cost, latency, KV cache, and reasoning quality. As long reasoning and agentic workflows get longer, token

Half the time I can’t understand what my agent is trying to explain. Am I dumb, or does it just not know how to speak human?

Knowing only one policy can be even less useful than knowing nothing at all. Training a value/Q model only on the current policy right before RL feels more like warmup, not pretraining. The policy changes, and V/Q changes. Real value pretraining should learn signals that transfer

A paper that transcends its era

We Actually Solved the “Ma Jiaqi Problem” Long Ago MiniMax’s recent analysis of the “Ma Jiaqi problem” is quite interesting. Simply put, the model does not fail because it does not know who Ma Jiaqi is. It can answer questions about his group, identity, and debut information,