Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking? (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing jailbreak attacks fail against reasoning models enhanced by Chain-of-Thought (CoT) reasoning. |
| Approach: | They propose a jailbreak method that uses Chain-of-Thought reasoning to reduce harmfulness from jailbreaking. |
| Outcome: | The proposed jailbreak method performs well against open AI models and deepseek-R1 reasoning models. |
Similar Papers
Focus on Your Question! Interpreting and Mitigating Toxic CoT Problems in Commonsense Reasoning (2024.acl-long)
Copied to clipboard
| Challenge: | Large language models exhibit high-level commonsense reasoning abilities, especially with enhancement methods like Chain-of-Thought (CoT). |
| Approach: | They propose a chain-of-thought-like method to elicit models' potential abilities to generate rationales and answers that are based on attribution tracing and causal tracers to probe the internal working mechanism of the LLM. |
| Outcome: | The proposed method eliminates Toxic CoT problems and improves the model’s overall commonsense reasoning performance by 5.5%. |
Preemptive Answer “Attacks” on Chain-of-Thought Reasoning (2024.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) have impressive reasoning capabilities when coupled with Chain-of-Thought (CoT) prompting. |
| Approach: | They propose a scenario where the LLM obtains an answer before engaging in reasoning, and propose two measures to bolster the robustness of this approach. |
| Outcome: | The proposed model significantly impairs its reasoning capability across various CoT methods and a broad spectrum of datasets. |
Analysing Chain of Thought Dynamics: Active Guidance or Unfaithful Post-hoc Rationalisation? (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent work has demonstrated that using chain of thought (CoT) on soft-reasoning tasks can yield limited or even negative performance gains. |
| Approach: | They investigate how chain of thought (CoT) is used in soft-reasoning tasks across instruction-tuned, reasoning and reasoning-distilled models. |
| Outcome: | The proposed model can steer predictions without faithfully reflecting reasoning, indicating a disconnect between CoT influence and faithfulness. |
Critical-CoT: A Robust Defense Framework against Reasoning-Level Backdoor Attacks in Large Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are vulnerable to backdoors that use long-form reasoning to generate a specific word, choice, or class. |
| Approach: | They propose a mechanism that allows LLMs to develop critical thinking behaviors and detect backdoors by a two-stage fine-tuning. |
| Outcome: | The proposed mechanism exhibits strong cross-domain and cross-task generalization. |
Understanding Jailbreak Success: A Study of Latent Space Dynamics in Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Emerging jailbreaking techniques can still elicit unsafe outputs, presenting an ongoing challenge for model alignment. |
| Approach: | They propose to extract a jailbreak vector from a single class of jailbreaks that works to mitigate jailbreak effectiveness from other, semantically-dissimilar classes. |
| Outcome: | The proposed jailbreak vectors show that they reduce harmfulness in most models, and that they are similar in geometric and effect similarity. |
Thinking Traps in Long Chain-of-Thought: A Measurable Study and Trap-Aware Adaptive Restart (2026.findings-acl)
Copied to clipboard
null Chenkang, Fan Yu, Junjie Nian, Sihan Zhao, Zhuoka Feng, Zijun Yao, Wang Heng, Yu Minshen, Yixin Cao
| Challenge: | Experiments show that extended generation does not guarantee correctness . a recurring pattern in Long-CoT failures is a problem for large reasoning models . |
| Approach: | They propose a test-time control framework that truncates the trajectory before the trap segment and adaptively restarts decoding. |
| Outcome: | Experiments show that TAAR improves reasoning performance without fine-tuning model parameters. |
When Models Outthink Their Safety: Unveiling and Mitigating Self-Jailbreak in Large Reasoning Models (2026.findings-acl)
Copied to clipboard
Yingzhi Mao, Chunkang Zhang, Junxiang Wang, Xinyan Guan, Boxi Cao, Yaojie Lu, Hongyu Lin, Xianpei Han, Le Sun
| Challenge: | Existing methods often apply coarse-grained constraints over entire reasoning trajectories . Existing approaches often apply unsafe constraints, causing unsafe outputs . |
| Approach: | They propose a trajectory-level training framework that mitigates Self-Jailbreak . they propose 'chain-of-guardrail' to mitigate self-jailbreak by targeting step-level interventions . |
| Outcome: | The proposed framework mitigates Self-Jailbreak by targeting step-level interventions while maintaining reasoning ability. |
Confusion is the Final Barrier: Rethinking Jailbreak Evaluation and Investigating the Real Misuse Threat of LLMs (2025.findings-emnlp)
Copied to clipboard
Yu Yan, Sheng Sun, Zhe Wang, Yijun Lin, Zenghao Duan, Zhifei Zheng, Min Liu, Zhiyi Yin, Jianping Zhang
| Challenge: | Large Language Models have been developed to deal with real-world crimes, but it remains unclear whether they internalize authentic knowledge or are forced to simulate toxic language patterns. |
| Approach: | They construct knowledge-intensive Q&A to investigate misuse threats of Large Language Models in terms of dangerous knowledge possession, harmful task planning utility, and harmfulness judgment robustness. |
| Outcome: | The findings raise concerns that jailbreak success is often attributable to a hallucination loop between jailbroken LLM and judger LLM . |
Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens (2026.findings-acl)
Copied to clipboard
Chengshuai Zhao, Zhen Tan, Pingchuan Ma, Dawei Li, Bohan Jiang, Yancheng Wang, Yingzhen Yang, Huan Liu
| Challenge: | Chain-of-Thought (CoT) prompting has been shown to be effective in eliciting structured reasoning from large language models (LLMs). |
| Approach: | They propose a data distribution lens to understand when and why CoT reasoning fails . they propose 'data-based' training that trains LLMs from scratch . |
| Outcome: | The proposed model enables models to generate reasoning trajectories that approximate those observed during training. |
Chain-of-Probe: Examining the Necessity and Accuracy of CoT Step-by-Step (2025.findings-naacl)
Copied to clipboard
Zezhong Wang, Xingshan Zeng, Weiwen Liu, Yufei Wang, Liangyou Li, Yasheng Wang, Lifeng Shang, Xin Jiang, Qun Liu, Kam-Fai Wong
| Challenge: | Current research found the issue of Early Answering in large language models where the models already have an answer before generating the Chain-of-Thought (CoT). |
| Approach: | They propose a method to probe changes in confidence during the model’s reasoning and prioritize answers with correct reasoning among multiple candidates. |
| Outcome: | The proposed method reveals that in a significant number of question-answer cases, CoT appears to be unnecessary and this necessity correlates with the simplicity of the task, defined by the reasoning steps required. |