REVIEW 11 cited by
SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities
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
SafeChain: Safety of Language Models with Long Chain-of-Thought Reasoning Capabilities
read the original abstract
Emerging large reasoning models (LRMs), such as DeepSeek-R1 models, leverage long chain-of-thought (CoT) reasoning to generate structured intermediate steps, enhancing their reasoning capabilities. However, long CoT does not inherently guarantee safe outputs, potentially leading to harmful consequences such as the introduction of security vulnerabilities in code or the spread of misinformation. Current research on large language model (LLM) safety usually focuses on short-answer responses, overlooking the long CoT style outputs of LRMs. To bridge this gap, we conduct a systematic study of LRM safety. First, we investigate safety evaluators calibrated against human annotations. Using our newly developed metrics, we thoroughly assess the safety of 12 state-of-the-art LRMs on StrongReject and WildJailbreak datasets. Our results show that LRMs are not safe compared to their reasoning advance. Further, we perform a fine-grained analysis of the reasoning trace and final answer. We find that three decoding strategies-ZeroThink, LessThink, and MoreThink-can improve model safety without additional training. However, these strategies either use constrained reasoning traces or incur high inference costs. To better strengthen LRM safety, we introduce SafeChain, the first-of-its-kind safety training dataset in CoT style. We fine-tune two LRMs with SafeChain, showing that it not only enhances model safety but also preserves performance across 6 reasoning benchmarks.
Forward citations
Cited by 11 Pith papers
-
Safety Under Scaffolding: How Evaluation Conditions Shape Measured Safety
Map-reduce scaffolding degrades measured safety mainly by stripping multiple-choice options (40–89% of the loss is format conversion); scaffold architecture explains only 0.4% of variance and composite safety scores h...
-
Fragile Thoughts: How Large Language Models Handle Chain-of-Thought Perturbations
LLMs show heterogeneous robustness to five types of chain-of-thought perturbations, with MathError causing 50-60% accuracy loss in small models but scaling benefits, UnitConversion remaining hard across sizes, and Ext...
-
When Models Outthink Their Safety: Unveiling and Mitigating Self-Jailbreak in Large Reasoning Models
Large Reasoning Models override their own initial safety recognition during multi-step reasoning in a failure mode called Self-Jailbreak, which Chain-of-Guardrail mitigates through targeted trajectory-level step inter...
-
Addressing Over-Refusal in LLMs with Competing Rewards
SEAR trains one LLM via adversarial process rewards to explore harmful reasoning paths but flip to safe outputs, reducing over-refusal while preserving safety.
-
MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety
MultiBreak is a large diverse multi-turn jailbreak benchmark that achieves substantially higher attack success rates on LLMs than prior datasets and reveals topic-specific vulnerabilities in multi-turn settings.
-
Reasoning Structure Matters for Safety Alignment of Reasoning Models
Changing the internal reasoning structure of large reasoning models through simple supervised fine-tuning on 1K examples produces strong safety alignment that generalizes across tasks and languages.
-
Selective Forgetting for Large Reasoning Models
FRUL uses multi-LLM RAG to isolate and replace sensitive CoT segments, then applies a suppress-and-reinforce loss so LRMs forget targeted knowledge while retaining structured reasoning.
-
ReasoningGuard: Safeguarding Large Reasoning Models with Inference-time Safety Aha Moments
ReasoningGuard is an inference-time method that uses attention mechanisms to inject safety aha moments and scaling sampling to defend large reasoning models against jailbreak attacks.
-
Position: The Hidden Costs and Measurement Gaps of Reinforcement Learning with Verifiable Rewards
The paper identifies confounds in RLVR evaluations that inflate apparent gains and proposes a minimum standard for budget-matched, contamination-aware assessment with calibration tracking.
-
Phi-4-reasoning Technical Report
A 14B reasoning model trained via supervised fine-tuning on selected prompts and o3-mini traces, plus outcome RL, outperforms larger open models like DeepSeek-R1-Distill-Llama-70B on math, coding, planning and related...
-
A Survey of Scaling in Large Language Model Reasoning
A survey categorizing scaling in LLM reasoning across input size, steps, rounds, training, and future directions, noting that scaling can negatively affect performance.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.