REVIEW 5 cited by
Controllable Safety Alignment: Inference-Time Adaptation to Diverse Safety Requirements
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The current paradigm for safety alignment of large language models (LLMs) follows a one-size-fits-all approach: the model refuses to interact with any content deemed unsafe by the model provider. This approach lacks flexibility in the face of varying social norms across cultures and regions. In addition, users may have diverse safety needs, making a model with static safety standards too restrictive to be useful, as well as too costly to be re-aligned. We propose Controllable Safety Alignment (CoSA), a framework designed to adapt models to diverse safety requirements without re-training. Instead of aligning a fixed model, we align models to follow safety configs -- free-form natural language descriptions of the desired safety behaviors -- that are provided as part of the system prompt. To adjust model safety behavior, authorized users only need to modify such safety configs at inference time. To enable that, we propose CoSAlign, a data-centric method for aligning LLMs to easily adapt to diverse safety configs. Furthermore, we devise a novel controllability evaluation protocol that considers both helpfulness and configured safety, summarizing them into CoSA-Score, and construct CoSApien, a human-authored benchmark that consists of real-world LLM use cases with diverse safety requirements and corresponding evaluation prompts. We show that CoSAlign leads to substantial gains of controllability over strong baselines including in-context alignment. Our framework encourages better representation and adaptation to pluralistic human values in LLMs, and thereby increasing their practicality.
Forward citations
Cited by 5 Pith papers
-
Reasoning Up the Instruction Ladder for Controllable Language Models
RLVR on ~7K verifiable system/user conflict examples teaches LLMs to prioritize higher-priority instructions, improving instruction-hierarchy and safety benchmarks.
-
NeuronTune: Fine-Grained Neuron Modulation for Balanced Safety-Utility Alignment in LLMs
NeuronTune identifies sparse safety and utility neurons via attack-aware attribution, optimizes per-neuron scaling factors with MAML, and reports a better safety-utility balance than layer-wise alignment methods.
-
"I Cannot Write This Because It Violates Our Content Policy": Understanding Content Moderation Policies and User Experiences in Generative AI Products
GAI tools' content moderation policies are comprehensive in scope but thin on user reporting and appeals, and Reddit users report frequent frustration with opaque moderation decisions.
-
Personalized Constitutionally-Aligned Agentic Superego: Secure AI Behavior Aligned to Diverse Human Values
An external 'superego' module that filters agentic AI plans against user-selected 'constitutions' plus a universal safety floor is reported to cut harmful outputs by up to 98% on safety benchmarks.
-
A Survey on Training-free Alignment of Large Language Models
A survey that catalogs and categorizes training-free LLM alignment methods into pre-decoding, in-decoding, and post-decoding, with a limited experimental comparison on one model.
Discussion (0). Sign in to comment.