Token cues unlock latent reasoning behavior in base models

Tue, 06 Oct 2026 16:01:24 GMT

Researchers show that fixed response-opening token cues can unlock reasoning behavior already associated with training data. The cue ".\n\nOkay" raises OLMo-3-7B from 42% to 78% on MATH-500, while "Alright," raises Qwen3-14B from 72% to 87%. Causal data interventions can create or remove a cue, and related effects also appear in refusal behavior. Code is public, but results are author-reported.

Why read: Token cues unlock latent base-model capability at negligible inference cost while exposing a new behavior-control surface.

Recommendation: TRIAL

Sources: Base Models Can Reason By Taking a Cue From Training Data · sophicle/cues

MATRIX · Token Cues

Token cues unlock latent reasoning behavior in base models