English
Related papers

Related papers: Epistemic Traps: Rational Misalignment Driven by M…

200 papers

Specification gaming under Reinforcement Learning (RL) is known to cause LLMs to develop sycophantic, manipulative, or deceptive behavior, yet the conditions under which this occurs remain unclear. We train 11 instruction-tuned LLMs…

Machine Learning · Computer Science 2026-04-15 Leon Eshuijs , Shihan Wang , Antske Fokkens

This chapter develops a unified framework for studying misspecified learning situations in which agents optimize and update beliefs within an incorrect model of their environment. We review the statistical foundations of learning from…

Theoretical Economics · Economics 2026-01-16 Ignacio Esponda , Demian Pouzo

Classical reinforcement learning assumes the agent interacts with a fixed environment whose behavior does not depend on the agent's policy. This assumption breaks down in non-realizable settings where other actors might anticipate the…

In bandit settings, optimizing long-term regret metrics requires exploration, which corresponds to sometimes taking myopically sub-optimal actions. When a long-lived principal merely recommends actions to be executed by a sequence of…

Computer Science and Game Theory · Computer Science 2026-02-25 Ramya Ramalingam , Osbert Bastani , Aaron Roth

Reinforcement learning (RL) systems can be complex and non-interpretable, making it challenging for non-AI experts to understand or intervene in their decisions. This is due in part to the sequential nature of RL in which actions are chosen…

Artificial Intelligence · Computer Science 2025-04-16 Amal Alabdulkarim , Madhuri Singh , Gennie Mansi , Kaely Hall , Upol Ehsan , Mark O. Riedl

Artificial Intelligence (AI) systems are increasingly placed in positions where their decisions have real consequences, e.g., moderating online spaces, conducting research, and advising on policy. Ensuring they operate in a safe and…

Artificial Intelligence · Computer Science 2025-05-09 Joel Z. Leibo , Alexander Sasha Vezhnevets , William A. Cunningham , Sébastien Krier , Manfred Diaz , Simon Osindero

The increasing adoption of Reinforcement Learning in safety-critical systems domains such as autonomous vehicles, health, and aviation raises the need for ensuring their safety. Existing safety mechanisms such as adversarial training,…

Machine Learning · Computer Science 2021-11-11 Paulina Stevia Nouwou Mindom , Amin Nikanjam , Foutse Khomh , John Mullins

General Alignment has improved average-case helpfulness and safety, but current alignment practice still rewards confident, single-turn responses. The problem is not only that models fail on edge cases; it is that current evaluation makes…

Computation and Language · Computer Science 2026-05-19 Han Bao , Yue Huang , Xiaoda Wang , Zheyuan Zhang , Yujun Zhou , Carl Yang , Xiangliang Zhang , Yanfang Ye

Automating scientific computing workflows requires more than generating executable code: autonomous systems must also select appropriate computational strategies, implement them faithfully, and ensure that the resulting outcomes remain…

Artificial Intelligence · Computer Science 2026-05-29 Geremy Loachamín-Suntaxi , Robert Lazar , Dimitrios G. Giovanis , Ioannis G. Kevrekidis , Eleni D. Koronaki

We study Markov decision problems where the agent does not know the transition probability function mapping current states and actions to future states. The agent has a prior belief over a set of possible transition functions and updates…

Economics · Quantitative Finance 2018-01-04 Ignacio Esponda , Demian Pouzo

The integration of Large Language Models (LLMs) into scientific discovery is currently hindered by the Implicit Context problem, where governing equations extracted from literature contain invisible thermodynamic assumptions (e.g.,…

Databases · Computer Science 2026-03-11 Yue Wua , Tianhao Su , Rui Hu , Mingchuan Zhao , Shunbo Hu , Deng Pan , Jizhong Huang

Large language model (LLM) agents increasingly rely on external tools (file operations, API calls, database transactions) to autonomously complete complex multi-step tasks. Practitioners deploy defense-trained models to protect against…

Cryptography and Security · Computer Science 2026-03-23 Shawn Li , Yue Zhao

We stress-tested 16 leading models from multiple developers in hypothetical corporate environments to identify potentially risky agentic behaviors before they cause real harm. In the scenarios, we allowed models to autonomously send emails…

Cryptography and Security · Computer Science 2025-10-17 Aengus Lynch , Benjamin Wright , Caleb Larson , Stuart J. Ritchie , Soren Mindermann , Evan Hubinger , Ethan Perez , Kevin Troy

Human decision-making in real-life deviates significantly from the optimal decisions made by fully rational agents, primarily due to computational limitations or psychological biases. While existing studies in behavioral finance have…

Artificial Intelligence · Computer Science 2024-03-12 Penghang Liu , Kshama Dwarakanath , Svitlana S Vyetrenko , Tucker Balch

Motivated by loss of control risks from misaligned AI systems, we develop and apply methods for measuring language models' propensity for unsanctioned behaviour. We contribute three methodological improvements: analysing effects of changes…

Artificial Intelligence · Computer Science 2026-04-24 Olli Järviniemi , Oliver Makins , Jacob Merizian , Robert Kirk , Ben Millwood

We study misspecified Bayesian learning in principal-agent relationships, where an agent is assessed by an evaluator and rewarded by the market. The agent's outcome depends on their innate ability, costly effort -- whose effectiveness is…

Theoretical Economics · Economics 2025-12-02 Federico Echenique , Anqi Li

Recent studies on the safety alignment of large language models (LLMs) have revealed that existing approaches often operate superficially, leaving models vulnerable to various adversarial attacks. Despite their significance, these studies…

Cryptography and Security · Computer Science 2025-06-02 Jianwei Li , Jung-Eun Kim

Standard models of bounded rationality typically assume agents either possess accurate knowledge of the population's reasoning abilities (Cognitive Hierarchy) or hold dogmatic, degenerate beliefs (Level-$k$). We introduce the ``Connected…

Computer Science and Game Theory · Computer Science 2026-02-13 Raman Ebrahimi , Sepehr Ilami , Babak Heydari , Isabel Trevino , Massimo Franceschetti

Autonomous agents powered by LLMs and Retrieval-Augmented Generation (RAG) are proficient consumers of digital content but remain unidirectional, a limitation we term epistemic asymmetry. This isolation leads to redundant reasoning and…

Artificial Intelligence · Computer Science 2025-12-25 Zan-Kai Chong , Hiroyuki Ohsaki , Bryan Ng

Linearly transforming stimulus representations of deep neural networks yields high-performing models of behavioral and neural responses to complex stimuli. But does the test accuracy of such predictions identify genuine representational…

Neurons and Cognition · Quantitative Biology 2026-01-05 Itamar Avitan , Tal Golan