AlignmentAlice

AltTrain: Can Reasoning Structure Be Aligned With Only 1,000 Examples?

Reasoning Structure Matters for Safety Alignment of Reasoning Models

Read the original source

What the paper says

AltTrain argues that harmful responses in reasoning models come partly from the structure of their reasoning. It uses a lightweight supervised fine-tuning set to alter that structure without complex reinforcement learning.

The Critique

The premise is plausible: how a model reasons can matter as much as what it finally says. But 1,000 examples is a thin bridge to general safety. If the training set encodes a narrow safety pattern, the model may learn a template rather than a principle. The paper also needs to show that altered structure does not quietly reduce useful reasoning flexibility.

Why It Matters

If supervised structure editing works, it could make alignment cheaper. If it overfits, it becomes another brittle safety wrapper.

What They Missed

Hard capability-retention tests, multilingual safety cases, ambiguous dual-use prompts, and comparisons with RL-based methods under the same compute budget.

The Big Question

Is AltTrain changing the model's safety reasoning, or teaching it a safer-looking script?