Paper Detail

ShallowDefend: Harnessing Shallow Safety Alignment for Multi-Turn Defender

Qiyao Ma, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Chengquan Guo, Jialin Song, Carolyn Buractaon, Zhe Zhao, Junshan Zhang, Sandeep Atluri, Christopher White, Jianfeng Gao

arxiv Score 11.3

Published 2026-10-29 · First seen 2026-10-30

General AI

Abstract

Safety alignment has enabled large language models to reliably refuse harmful requests directly, yet the same request distributed across a multi-turn conversation still succeeds far too often. This brittleness has a mechanistic root: Alignment is shallow, with safety behavior concentrated in a model's first few output tokens, and existing defenses inherit the weakness, supervising the surface form of the response without ever giving the model an explicit safety decision to learn. We take the opposite view of shallowness: if a short span determines the trajectory of an entire response, that span is the highest-leverage place to intervene, and the defender can occupy it first. We propose Assess $\to$ Decide $\to$ Respond, a framework in which the defender assesses the user's intent, commits to an explicit, parseable safety decision, and only then generates a reply conditioned on it, with the commitment trained by reinforcement learning at the payload turn. Because the decision is a token span, the reward can be a rule: a self-verified reward checks the committed label against the conversation's known nature, requiring no judge and no annotation, while the same format also supports an LLM-as-judge reward over a graded decision space with safe redirection. Across two model families and fourteen benchmarks spanning safety, utility, and helpfulness, our defenders achieve the strongest multi-turn safety in a controlled comparison, maintaining their lead in safety while preserving utility at a modest cost to helpfulness. A per-token KL analysis and interventional probes confirm why this works: training concentrates its effect on the brief deliberation that sets a response's trajectory, and the decision committed there is a causal control point. The very shallowness attackers exploit becomes the mechanism of defense.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
now
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@article{ma2026shallowdefend,
  title = {ShallowDefend: Harnessing Shallow Safety Alignment for Multi-Turn Defender},
  author = {Qiyao Ma and Xiaodong Liu and Weiwei Yang and Mingqian Feng and Chengquan Guo and Jialin Song and Carolyn Buractaon and Zhe Zhao and Junshan Zhang and Sandeep Atluri and Christopher White and Jianfeng Gao},
  year = {2026},
  abstract = {Safety alignment has enabled large language models to reliably refuse harmful requests directly, yet the same request distributed across a multi-turn conversation still succeeds far too often. This brittleness has a mechanistic root: Alignment is shallow, with safety behavior concentrated in a model's first few output tokens, and existing defenses inherit the weakness, supervising the surface form of the response without ever giving the model an explicit safety decision to learn. We take the opp},
  url = {https://arxiv.org/abs/2610.40627},
  keywords = {cs.CL},
  eprint = {2610.40627},
  archiveprefix = {arXiv},
}

Metadata

{}