Small Decision Model
A language model with a parameter count between several hundred million and several billion (for example, 1B-3B). It has short reasoning paths, low memory usage and low per-call cost, making it suitable for structured decision tasks such as intent recognition, tool selection, parameter generation, classification and slot filling, though its ability to generalize to unseen intents is weaker than that of large models.
Timeline (last 90 days)
The applicable boundary of a small decision model depends on the degree of task determinism: when the tool set is fixed, the business process is enumerable, and user intents can be summarized into a finite number of categories, a decision model can cover most cases with a very small number of parameters; once it enters domains that are ambiguous, polysemous, and require a large amount of world knowledge, the decision model may lack sufficient semantic understanding capability.
If tools change frequently and processes are often adjusted, the maintenance cost of the decision model may offset its savings at inference time.
All Facts (2)
The applicable boundary of a small decision model depends on the degree of task determinism: when the tool set is fixed, the business process is enumerable, and user intents can be summarized into a finite number of categories, a decision model can cover most cases with a very small number of parameters; once it enters domains that are ambiguous, polysemous, and require a large amount of world knowledge, the decision model may lack sufficient semantic understanding capability.
50%UnverifiedIf tools change frequently and processes are often adjusted, the maintenance cost of the decision model may offset its savings at inference time.
50%