Alignment with Reflect-Reasoning: Learning How To Follow Principles
Paper · https://arxiv.org/abs/2611.17398
arXiv:2611.17398v1 Announce Type: new Abstract: The appropriate governing norms and principles for the alignment of Large Language Models (LLMs) often depend on cultural context and deployment application. However, most alignment techniques focus on training models to conform to human preferences or constitutional principles specified at training time and subsequently fixed. In practice, many deployers rely on system prompting to implement contextual norms and guardrails at inference time. Prior work has shown that post-generation reflective reasoning with critique and revision can improve conformance compared to simple system prompting, but such methods are limited to the reasoning abilities inherent in the trained model. We introduce Reflect-Reasoning, a training method that teaches models how to reason about alignment to principles, rather than simply which principles to follow. Across five models, Reflect-Reasoning consistently outperforms both system prompting and explicit prompting strategies for in-context reflective reasoning. These improvements extend and generalize to entirely out-of-distribution domains of medical and financial principles. We additionally demonstrate that Reflect-Reasoning reduces the occurrence of rare but serious principle violations. Finally, we show that Reflect-Reasoning does not erode the safety engineering of a model and is robust against jailbreaking attempts using adversarial constitutions. Overall, our results demonstrate that Reflect-Reasoning is a scalable method for training constitutional alignment as a generalizable reasoning capability, enabling models to adapt to novel principles and produce transparent reasoning traces without additional fine-tuning.
First seen by Oriel on .
AI summary
The paper introduces Reflect-Reasoning, a training method that teaches LLMs to reason about alignment principles and adapt to novel constitutional requirements without additional fine-tuning.
- Reflect-Reasoning consistently outperforms system prompting and explicit prompting for in-context reflective reasoning across five models.
- Its improvements generalize to out-of-distribution medical and financial principles, rather than remaining tied to training-time constitutions.
- The method reduces rare but serious principle violations while preserving the model’s existing safety engineering.
- Reflect-Reasoning remains robust against jailbreaking attempts using adversarial constitutions and produces transparent reasoning traces.
#llm alignment #reflective reasoning #constitutional ai #jailbreaking
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| arXiv AI Papers | oai:arXiv.org:2611.17398v1 | https://arxiv.org/abs/2611.17398 Read discussion |