Copertina

Exploration Hacking: Can LLMs Learn to Resist RL Training?

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Jang, Eyon, Falck, Damon, Braun, Joschka, Kirch, Nathalie, Menon, Achu, Moodley, Perusha, Emmons, Scott, Zimmermann, Roland S., Lindner, David
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning Computation and Language
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Stress-Testing Alignment Audits With Prompt-Level Strategic Deception
di: Daniels, Oliver, et al.
Pubblicazione: (2026)

Inherited Goal Drift: Contextual Pressure Can Undermine Agentic Goals
di: Menon, Achyutha, et al.
Pubblicazione: (2026)

A Pragmatic Way to Measure Chain-of-Thought Monitorability
di: Emmons, Scott, et al.
Pubblicazione: (2025)

Understanding Unreliability of Steering Vectors in Language Models: Geometric Predictors and the Limits of Linear Approximations
di: Braun, Joschka
Pubblicazione: (2026)

Asymmetric Goal Drift in Coding Agents Under Value Conflict
di: Saebo, Magnus, et al.
Pubblicazione: (2026)

Early Signs of Steganographic Capabilities in Frontier LLMs
di: Zolkowski, Artur, et al.
Pubblicazione: (2025)

MONA: Myopic Optimization with Non-myopic Approval Can Mitigate Multi-step Reward Hacking
di: Farquhar, Sebastian, et al.
Pubblicazione: (2025)

Multi-State-Action Tokenisation in Decision Transformers for Multi-Discrete Action Spaces
di: Moodley, Perusha, et al.
Pubblicazione: (2024)

Aligned, Orthogonal or In-conflict: When can we safely optimize Chain-of-Thought?
di: Kaufmann, Max, et al.
Pubblicazione: (2026)

Revisiting Safe Exploration in Safe Reinforcement learning
di: Eckel, David, et al.
Pubblicazione: (2024)

What Features in Prompts Jailbreak LLMs? Investigating the Mechanisms Behind Attacks
di: Kirch, Nathalie, et al.
Pubblicazione: (2024)

Neural Chameleons: Language Models Can Learn to Hide Their Thoughts from Unseen Activation Monitors
di: McGuinness, Max, et al.
Pubblicazione: (2025)

TRIAGE: Ethical Benchmarking of AI Models Through Mass Casualty Simulations
di: Kirch, Nathalie Maria, et al.
Pubblicazione: (2024)

Can LLMs Hack Enterprise Networks? -- Replicated Computational Results (RCR) Report
di: Happe, Andreas, et al.
Pubblicazione: (2026)

Welcome First--Books Later; The Service Center Branch, Richmond Public Library, December 1967 - June 1971.
di: Emmons, Karen
Pubblicazione: (1971)

She's Practiced What She Teaches
di: Emmons, Julia
Pubblicazione: (1976)

The Impact of Off-Policy Training Data on Probe Generalisation
di: Kirch, Nathalie, et al.
Pubblicazione: (2025)

Natural Emergent Misalignment from Reward Hacking in Production RL
di: MacDiarmid, Monte, et al.
Pubblicazione: (2025)

Logit Reweighting for Topic-Focused Summarization
di: Braun, Joschka, et al.
Pubblicazione: (2025)

Beyond Multiple Choice: Evaluating Steering Vectors for Summarization
di: Braun, Joschka, et al.
Pubblicazione: (2025)

Land of 10,000 Publishers: Minnesota Children's Book Publishing
di: Kirch, Claire
Pubblicazione: (2008)

Financial constraints and the interdependence of corporate ﬁnancial decisions A cross-country study
di: Guilherme Kirch
Pubblicazione: (2020)

School of Reward Hacks: Hacking harmless tasks generalizes to misaligned behavior in LLMs
di: Taylor, Mia, et al.
Pubblicazione: (2025)

Imbalanced Gradients in RL Post-Training of Multi-Task LLMs
di: Wu, Runzhe, et al.
Pubblicazione: (2025)

Rapid Integration of LLMs in Healthcare Raises Ethical Concerns: An Investigation into Deceptive Patterns in Social Robots
di: Ranisch, Robert, et al.
Pubblicazione: (2024)

The Ethics of ChatGPT in Medicine and Healthcare: A Systematic Review on Large Language Models (LLMs)
di: Haltaufderheide, Joschka, et al.
Pubblicazione: (2024)

Can LLMs Hack Enterprise Networks? Autonomous Assumed Breach Penetration-Testing Active Directory Networks
di: Happe, Andreas, et al.
Pubblicazione: (2025)

Efficient RL Training for LLMs with Experience Replay
di: Arnal, Charles, et al.
Pubblicazione: (2026)

We Can Imagine the Future, but Are We Equipped to Create It?
di: Jaggars, Damon E.
Pubblicazione: (2014)

Language Models Can Autonomously Hack and Self-Replicate
di: Air, Alena, et al.
Pubblicazione: (2026)

LLMs Gaming Verifiers: RLVR can Lead to Reward Hacking
di: Helff, Lukas, et al.
Pubblicazione: (2026)

Criança e adolescente: a problemática da adoção e posterior devolução às casas de acolhimento
di: Aline Taiane Kirch
Pubblicazione: (2014)

Knapsack RL: Unlocking Exploration of LLMs via Optimizing Budget Allocation
di: Li, Ziniu, et al.
Pubblicazione: (2025)

ALMANACS: A Simulatability Benchmark for Language Model Explainability
di: Mills, Edmund, et al.
Pubblicazione: (2023)

Observation Interference in Partially Observable Assistance Games
di: Emmons, Scott, et al.
Pubblicazione: (2024)

Image Hijacks: Adversarial Images can Control Generative Models at Runtime
di: Bailey, Luke, et al.
Pubblicazione: (2023)

LLMs Can Learn to Reason Via Off-Policy RL
di: Ritter, Daniel, et al.
Pubblicazione: (2026)

Understanding (Un)Reliability of Steering Vectors in Language Models
di: Braun, Joschka, et al.
Pubblicazione: (2025)

Vendor-Aware Industrial Agents: RAG-Enhanced LLMs for Secure On-Premise PLC Code Generation
di: Kersting, Joschka, et al.
Pubblicazione: (2025)

Agent^2 RL-Bench: Can LLM Agents Engineer Agentic RL Post-Training?
di: Chen, Wanyi, et al.
Pubblicazione: (2026)