Post-training hybrid reasoning models in NoThink mode has attracted growing interest as a way to improve performance while keeping inference fast.
However, these gains may draw on thinking behavior already accessible through the base model's Think mode.
We formulate this thinking leakage in a causal mediation framework and audit its contribution using bidirectional interventions along a simple base-derived activation direction.
Findings
Across three models and three post-training methods on competition math benchmarks, we find that leakage is real, causal, and substantial:
- Behavioral and representational analyses reveal shifts toward Think.
- Steering the base model along this direction reproduces most of the post-training accuracy gain.
- Counter-steering a checkpoint removes a substantial share of what it gains.
Across nine aligned checkpoints with positive NoThink gains, the resulting leakage ratio ranges from 42% to 79%.
These interventions support a substantial causal contribution of thinking leakage.
Our findings show that a post-training method's apparent advantage can therefore reflect greater drift toward Think, obscuring whether it improves capability within NoThink or more effectively re-invokes existing Think behavior.