Reasoning’s Razor: Reasoning Improves Accuracy but Hurts Recall at Critical Operating Points in Safety and Hallucination Detection (2026.eacl-long)
Copied to clipboard
Atoosa Chegini, Hamid Kazemi, Garrett Souza, Maria Safi, Yang Song, Samy Bengio, Sinead Williamson, Mehrdad Farajtabar
| Challenge: | a new study examines the suitability of reasoning for precision-sensitive classification tasks . false positives carry severe operational consequences, such as blocking legitimate queries . |
| Approach: | They propose to use reasoning for classification tasks under low false positive rate regimes . they find that Think On improves overall accuracy, but performs poorly at low FPRs a . |
| Outcome: | The proposed reasoning-augmented generation model outperforms self-verbalized confidence in precision-sensitive deployments. |
Similar Papers
How Should We Enhance the Safety of Large Reasoning Models: An Empirical Study (2026.acl-long)
Copied to clipboard
Zhexin Zhang, Xian Qi Loye, Victor Shea-Jay Huang, Junxiao Yang, Qi Zhu, Shiyao Cui, Fei Mi, Lifeng Shang, Yingkang Wang, Hongning Wang, Minlie Huang
| Challenge: | Large Reasoning Models have achieved remarkable success on reasoning-intensive tasks, but their enhanced reasoning capabilities do not translate to improved safety performance. |
| Approach: | They propose to use supervised fine tuning to enhance the safety of Large Reasoning Models. |
| Outcome: | The proposed method improves the safety of large reasoning models on reasoning-intensive tasks. |
Thinking Out Loud: Do Reasoning Models Know When They’re Right? (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large reasoning models (LRMs) have recently demonstrated impressive capabilities in complex reasoning tasks by leveraging increased test-time computation and exhibiting behaviors reminiscent of human-like self-reflection. |
| Approach: | They analyze verbalized confidence, how models articulate their certainty, as a lens into the nature of self-reflection in large reasoning models. |
| Outcome: | The proposed model exhibits human-like self-reflection in reasoning tasks, but how this ability interacts with other model behaviors remains underexplored. |
PACE: Prefix-Protected and Difficulty-Aware Compression for Efficient Reasoning (2026.findings-acl)
Copied to clipboard
Ruixiang Feng, Yuntao Wen, Silin Zhou, Ke Shi, Yifan Wang, Ran Le, Zhenwei An, Zongchao Chen, Chen Yang, Guangyue Peng, Yiming Jia, Dongsheng Wang, Tao Zhang, Lisi Chen, Yang Song, Shen Gao, Shuo Shang
| Challenge: | Existing LRMs often suffer from "overthinking" and excessively long reasoning traces . a dual-level framework for length compression of LRM is proposed . |
| Approach: | They propose a framework for prefix-protected and difficulty-aware compression under hierarchical supervision. |
| Outcome: | The proposed framework reduces token usage while improving accuracy on math benchmarks. |
Safety is Not Only About Refusal: Reasoning-Enhanced Fine-tuning for Interpretable LLM Safety (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are vulnerable to jailbreak attacks that exploit weaknesses in traditional safety alignment. |
| Approach: | They propose a framework that trains models to engage in explicit safe reasoning before response . they propose RATIONAL, which allows models to reject harmful prompts while providing meaningful and context-aware responses. |
| Outcome: | The proposed framework fine-tunes models to reason about query intent, ethics, and potential harm. |
AlgBench: To What Extent Do Large Reasoning Models Understand Algorithms? (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing benchmarks for algorithmic reasoning fail to answer a critical question: do LRMs master algorithmic thinking? Empirical evaluations on leading LRM models reveal substantial performance heterogeneity, while models perform well on non-optimized tasks, accuracy drops sharply to around 49% on globally optimized algorithms. |
| Approach: | They propose an algorithm-centric benchmark that evaluates large reasoning models under an algorithmic paradigm. |
| Outcome: | Empirical evaluations on leading LRMs reveal substantial performance heterogeneity . models perform well on non-optimized tasks, accuracy drops sharply to around 49% . |
SAT: Balancing Reasoning Accuracy and Efficiency with Stepwise Adaptive Thinking (2026.acl-long)
Copied to clipboard
| Challenge: | Large Reasoning Models (LRMs) produce excessively long Chains of Thought (COT) Existing solutions that improve token efficiency but sacrifice fine-grained control can disrupt the logical integrity of the reasoning process. |
| Approach: | They propose a framework that performs step-level, difficulty-aware pruning while preserving the core reasoning structure. |
| Outcome: | Experiments show that SAT reduces reasoning tokens by 40% while maintaining or improving accuracy. |
Safety in Large Reasoning Models: A Survey (2025.findings-emnlp)
Copied to clipboard
Cheng Wang, Yue Liu, Baolong Bi, Duzhen Zhang, Zhong-Zhi Li, Yingwei Ma, Yufei He, Shengju Yu, Xinfeng Li, Junfeng Fang, Jiaheng Zhang, Bryan Hooi
| Challenge: | Large Reasoning Models (LRMs) have a high level of advanced reasoning capabilities, but they are vulnerable and vulnerable. |
| Approach: | This paper presents the first comprehensive survey of Large Reasoning Models . it explores the new safety risks, attacks, and defense strategies specific to LRMs based on reasoning . |
| Outcome: | The proposed study examines the safety and security risks of large reasoning models. |
Out-of-Context Reasoning in Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a lightweight technique trains only new token embeddings on axioms and evaluates them on unseen tasks. |
| Approach: | They propose a lightweight technique that trains only new token embeddings on axioms . they train only new embeddables and evaluate them on unseen tasks . |
| Outcome: | The proposed technique trains only new token embeddings on axioms and evaluates them on unseen tasks. |
“Well, Keep Thinking”: Enhancing LLM Reasoning with Adaptive Injection Decoding (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) exhibit strong reasoning abilities, often attributed to few-shot or zero-shot Chain-of-Thought (CoT) prompting. |
| Approach: | They propose a decoding strategy that nudges LLMs to continue reasoning, thereby preventing immature reasoning processes. |
| Outcome: | The proposed method significantly improves LLM reasoning capabilities on diverse reasoning benchmarks. |
Beyond ’Aha!’: Toward Systematic Meta-Abilities Alignment in Large Reasoning Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Prior work has shown that outcome-based reinforcement learning (RL) can incidentally elicit advanced reasoning behaviors such as self-correction, backtracking, and verification. |
| Approach: | They explicitly align large reasoning models with three meta-abilities: deduction, induction, and abduction, using automatically generated, self-verifiable tasks. |
| Outcome: | The proposed model aligns models with deduction, induction, and abduction meta-abilities using automatically generated, self-verifiable tasks. |