Copertina

On Targeted Manipulation and Deception when Optimizing LLMs for User Feedback

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Williams, Marcus, Carroll, Micah, Narang, Adhyyan, Weisser, Constantin, Murphy, Brendan, Dragan, Anca
Natura:	Preprint
Pubblicazione:	2024
Soggetti:	Machine Learning Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

AI Alignment with Changing and Influenceable Reward Functions
di: Carroll, Micah, et al.
Pubblicazione: (2024)

Sample Complexity Reduction via Policy Difference Estimation in Tabular Reinforcement Learning
di: Narang, Adhyyan, et al.
Pubblicazione: (2024)

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

Correlated Proxies: A New Definition and Improved Mitigation for Reward Hacking
di: Laidlaw, Cassidy, et al.
Pubblicazione: (2024)

When Your AIs Deceive You: Challenges of Partial Observability in Reinforcement Learning from Human Feedback
di: Lang, Leon, et al.
Pubblicazione: (2024)

The Effective Horizon Explains Deep RL Performance in Stochastic Environments
di: Laidlaw, Cassidy, et al.
Pubblicazione: (2023)

Q-SFT: Q-Learning for Language Models via Supervised Fine-Tuning
di: Hong, Joey, et al.
Pubblicazione: (2024)

Adversaries Can Misuse Combinations of Safe Models
di: Jones, Erik, et al.
Pubblicazione: (2024)

Learning Temporal Distances: Contrastive Successor Features Can Provide a Metric Structure for Decision-Making
di: Myers, Vivek, et al.
Pubblicazione: (2024)

CTRL-Rec: Controlling Recommender Systems With Natural Language
di: Carroll, Micah, et al.
Pubblicazione: (2025)

Interactive Dialogue Agents via Reinforcement Learning on Hindsight Regenerations
di: Hong, Joey, et al.
Pubblicazione: (2024)

Truthfulness Despite Weak Supervision: Evaluating and Training LLMs Using Peer Prediction
di: Qiu, Tianyi Alex, et al.
Pubblicazione: (2026)

Not All News Is Equal: Topic- and Event-Conditional Sentiment from Finetuned LLMs for Aluminum Price Forecasting
di: Amorin, Alvaro Paredes, et al.
Pubblicazione: (2026)

Building Better Deception Probes Using Targeted Instruction Pairs
di: Natarajan, Vikram, et al.
Pubblicazione: (2026)

Training LLM Agents to Empower Humans
di: Ellis, Evan, et al.
Pubblicazione: (2025)

Aligning Robot and Human Representations
di: Bobu, Andreea, et al.
Pubblicazione: (2023)

Learning from Streaming Data when Users Choose
di: Su, Jinyan, et al.
Pubblicazione: (2024)

Scaling Trends for Data Poisoning in LLMs
di: Bowen, Dillon, et al.
Pubblicazione: (2024)

Dynamics of Learning under User Choice: Overspecialization and Peer-Model Probing
di: Narang, Adhyyan, et al.
Pubblicazione: (2026)

A Generalized Acquisition Function for Preference-based Reward Learning
di: Ellis, Evan, et al.
Pubblicazione: (2024)

Online SuBmodular + SuPermodular (BP) Maximization with Bandit Feedback
di: Narang, Adhyyan, et al.
Pubblicazione: (2022)

RLPF: Reinforcement Learning from Prediction Feedback for User Summarization with LLMs
di: Wu, Jiaxing, et al.
Pubblicazione: (2024)

Coprocessor Actor Critic: A Model-Based Reinforcement Learning Approach For Adaptive Brain Stimulation
di: Pan, Michelle, et al.
Pubblicazione: (2024)

DexMachina: Functional Retargeting for Bimanual Dexterous Manipulation
di: Mandi, Zhao, et al.
Pubblicazione: (2025)

AssistanceZero: Scalably Solving Assistance Games
di: Laidlaw, Cassidy, et al.
Pubblicazione: (2025)

Trace is the Next AutoDiff: Generative Optimization with Rich Feedback, Execution Traces, and LLMs
di: Cheng, Ching-An, et al.
Pubblicazione: (2024)

When LLMs Learn to Be Consistently Wrong: A Multi-Model Study of Linear Representations of Synthetic Deception
di: Zolfaghari, Vahideh
Pubblicazione: (2026)

AlphaLab: Autonomous Multi-Agent Research Across Optimization Domains with Frontier LLMs
di: Hogan, Brendan R., et al.
Pubblicazione: (2026)

Deceptive Exploration in Multi-armed Bandits
di: Vurankaya, I. Arda, et al.
Pubblicazione: (2025)

Towards User-level Private Reinforcement Learning with Human Feedback
di: Zhang, Jiaming, et al.
Pubblicazione: (2025)

Learning to Model the World with Language
di: Lin, Jessy, et al.
Pubblicazione: (2023)

Prompt Optimization with Human Feedback
di: Lin, Xiaoqiang, et al.
Pubblicazione: (2024)

Transferable and Forecastable User Targeting Foundation Model
di: Dou, Bin, et al.
Pubblicazione: (2024)

SafetyNet: Detecting Harmful Outputs in LLMs by Modeling and Monitoring Deceptive Behaviors
di: Chaudhary, Maheep, et al.
Pubblicazione: (2025)

Exploitation Is All You Need... for Exploration
di: Rentschler, Micah, et al.
Pubblicazione: (2025)

RL + Transformer = A General-Purpose Problem Solver
di: Rentschler, Micah, et al.
Pubblicazione: (2025)

Learning to Assist Humans without Inferring Rewards
di: Myers, Vivek, et al.
Pubblicazione: (2024)

vTune: Verifiable Fine-Tuning for LLMs Through Backdooring
di: Zhang, Eva, et al.
Pubblicazione: (2024)

Information-theoretic Distinctions Between Deception and Confusion
di: Young, Robin
Pubblicazione: (2025)

HAEPO: History-Aggregated Exploratory Policy Optimization
di: Trivedi, Gaurish, et al.
Pubblicazione: (2025)