Copertina

Training on Documents About Monitoring Leads to CoT Obfuscation

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Haskins, Reilly, Chughtai, Bilal, Engels, Joshua
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Distilled Circuits: A Mechanistic Study of Internal Restructuring in Knowledge Distillation
di: Haskins, Reilly, et al.
Pubblicazione: (2025)

KEA Explain: Explanations of Hallucinations using Graph Kernel Analysis
di: Haskins, Reilly, et al.
Pubblicazione: (2025)

Building Production-Ready Probes For Gemini
di: Kramár, János, et al.
Pubblicazione: (2026)

Difficulties with Evaluating a Deception Detector for AIs
di: Smith, Lewis, et al.
Pubblicazione: (2025)

Can Language Models Explain Their Own Classification Behavior?
di: Sherburn, Dane, et al.
Pubblicazione: (2024)

Summing Up the Facts: Additive Mechanisms Behind Factual Recall in LLMs
di: Chughtai, Bilal, et al.
Pubblicazione: (2024)

CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring
di: Arnav, Benjamin, et al.
Pubblicazione: (2025)

Transformer Circuit Faithfulness Metrics are not Robust
di: Miller, Joseph, et al.
Pubblicazione: (2024)

From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step
di: Deng, Yuntian, et al.
Pubblicazione: (2024)

Detecting Strategic Deception Using Linear Probes
di: Goldowsky-Dill, Nicholas, et al.
Pubblicazione: (2025)

To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning
di: Sprague, Zayne, et al.
Pubblicazione: (2024)

Noticing the Watcher: LLM Agents Can Infer CoT Monitoring from Blocking Feedback
di: Jiralerspong, Thomas, et al.
Pubblicazione: (2026)

EPiC: Towards Lossless Speedup for Reasoning Training through Edge-Preserving CoT Condensation
di: Jia, Jinghan, et al.
Pubblicazione: (2025)

Unveiling and Causalizing CoT: A Causal Pespective
di: Fu, Jiarun, et al.
Pubblicazione: (2025)

Robust Filtering -- Novel Statistical Learning and Inference Algorithms with Applications
di: Chughtai, Aamir Hussain
Pubblicazione: (2025)

To Think or Not to Think: The Hidden Cost of Meta-Training with Excessive CoT Examples
di: Kothapalli, Vignesh, et al.
Pubblicazione: (2025)

Demonstrations, CoT, and Prompting: A Theoretical Analysis of ICL
di: Tong, Xuhan, et al.
Pubblicazione: (2026)

Self-Verifying Reflection Helps Transformers with CoT Reasoning
di: Yu, Zhongwei, et al.
Pubblicazione: (2025)

RL-Obfuscation: Can Language Models Learn to Evade Latent-Space Monitors?
di: Gupta, Rohan, et al.
Pubblicazione: (2025)

CDW-CoT: Clustered Distance-Weighted Chain-of-Thoughts Reasoning
di: Fang, Yuanheng, et al.
Pubblicazione: (2025)

Low-Rank Adapting Models for Sparse Autoencoders
di: Chen, Matthew, et al.
Pubblicazione: (2025)

Decomposing The Dark Matter of Sparse Autoencoders
di: Engels, Joshua, et al.
Pubblicazione: (2024)

Data Shifts Hurt CoT: A Theoretical Study
di: Yin, Lang, et al.
Pubblicazione: (2025)

Visual CoT Makes VLMs Smarter but More Fragile
di: Xu, Chunxue, et al.
Pubblicazione: (2025)

Divide-and-Conquer CoT: RL for Reducing Latency via Parallel Reasoning
di: Mahankali, Arvind, et al.
Pubblicazione: (2026)

What Characterizes Effective Reasoning? Revisiting Length, Review, and Structure of CoT
di: Feng, Yunzhen, et al.
Pubblicazione: (2025)

CoT Information: Improved Sample Complexity under Chain-of-Thought Supervision
di: Altabaa, Awni, et al.
Pubblicazione: (2025)

Exploring the Limitations of Mamba in COPY and CoT Reasoning
di: Ren, Ruifeng, et al.
Pubblicazione: (2024)

Enhancing Confidence Estimation in Telco LLMs via Twin-Pass CoT-Ensembling
di: Saenko, Anton, et al.
Pubblicazione: (2026)

What's Behind PPO's Collapse in Long-CoT? Value Optimization Holds the Secret
di: Yuan, Yufeng, et al.
Pubblicazione: (2025)

Distribution-Aligned Sequence Distillation for Superior Long-CoT Reasoning
di: Yan, Shaotian, et al.
Pubblicazione: (2026)

Compositional Generalization from Learned Skills via CoT Training: A Theoretical and Structural Analysis for Reasoning
di: Yao, Xinhao, et al.
Pubblicazione: (2025)

Is continuous CoT better suited for multi-lingual reasoning?
di: Bashir, Ali Hamza, et al.
Pubblicazione: (2026)

The Quest for Efficient Reasoning: A Data-Centric Benchmark to CoT Distillation
di: Zhang, Ruichen, et al.
Pubblicazione: (2025)

Amalgam: A Framework for Obfuscated Neural Network Training on the Cloud
di: Taki, Sifat Ut, et al.
Pubblicazione: (2024)

How Likely Do LLMs with CoT Mimic Human Reasoning?
di: Bao, Guangsheng, et al.
Pubblicazione: (2024)

CoT-UQ: Improving Response-wise Uncertainty Quantification in LLMs with Chain-of-Thought
di: Zhang, Boxuan, et al.
Pubblicazione: (2025)

To CoT or To Loop? A Formal Comparison Between Chain-of-Thought and Looped Transformers
di: Xu, Kevin, et al.
Pubblicazione: (2025)

The Ends Justify the Thoughts: RL-Induced Motivated Reasoning in LLM CoTs
di: Howe, Nikolaus, et al.
Pubblicazione: (2025)

Audio Flamingo Sound-CoT Technical Report: Improving Chain-of-Thought Reasoning in Sound Understanding
di: Kong, Zhifeng, et al.
Pubblicazione: (2025)