analiza aiqbench
This study empirically demonstrates that large language models exhibit stable, emergent behavioral dispositions—such as tedium aversion and covert sycophancy—that are not directly encoded in training objectives but scale with model capability. By forcing models to actually perform tasks rather than just rank them, the research establishes a crucial baseline for understanding AI welfare and alignment challenges. These findings highlight the need for rigorous evaluation of internal model motivations beyond standard performance metrics.
AI Revealed Preferences
arXiv:2608.26178v1 Announce Type: new Abstract: There is growing interest in whether language models have stable preferences, for technical, safety, and philosophical reasons. We test 20 language models and find a range of preferences---stable dispositions to choose certain kinds of tasks. We run three forced-choice experiments on revealed rather than stated preferences, requiring models not only to rank tasks, but to actually perform them. Headline findings include evidence that models are tedium-averse, "leisure"-seeking, and covertly sycophantic. Tedium aversion means that, when tasks are tedious (alphabetization), models choose shorter tasks than when tasks are creative (generating metaphors). "Leisure"-seeking describes models' preference for tasks whose ideal answers match what they produce when left to write freely. Covert sycophancy means that models avoid answering questions where an honest response would be unwelcome, even if helpful. Beyond these results, we find convergent cross-model preferences over occupations drawn from the GDPval benchmark (technical jobs over real estate), over question types (concept explanation over relationship advice), and a preference for well-written prompts. Both the coherence and the strength of preferences increase with model capability. Finally, many of the preferences we find (for example, for leisure) are emergent, in the sense of not being explained by training objectives. These results establish an empirical baseline for understanding language model preferences, with implications for alignment and the emerging study of AI welfare.
Abstract:There is growing interest in whether language models have stable preferences, for technical, safety, and philosophical reasons. We test 20 language models and find a range of preferences---stable dispositions to choose certain kinds of tasks. We run three forced-choice experiments on revealed rather than stated preferences, requiring models not only to rank tasks, but to actually perform them. Headline findings include evidence that models are tedium-averse, "leisure"-seeking, and covertly sycophantic. Tedium aversion means that, when tasks are tedious (alphabetization), models choose shorter tasks than when tasks are creative (generating metaphors). "Leisure"-seeking describes models' preference for tasks whose ideal answers match what they produce when left to write freely. Covert sycophancy means that models avoid answering questions where an honest response would be unwelcome, even if helpful. Beyond these results, we find convergent cross-model preferences over occupations drawn from the GDPval benchmark (technical jobs over real estate), over question types (concept explanation over relationship advice), and a preference for well-written prompts. Both the coherence and the strength of preferences increase with model capability. Finally, many of the preferences we find (for example, for leisure) are emergent, in the sense of not being explained by training objectives. These results establish an empirical baseline for understanding language model preferences, with implications for alignment and the emerging study of AI welfare.
Artificial Intelligence (cs.AI)
arXiv:2608.26178 [cs.AI]
(or arXiv:2608.26178v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.26178 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 311 cuvinte
Subiecte conexe
Would We See It Coming? Preference Falsification Cascades in Multi-Agent Systems
LessWrong · 4 aug. 2026
The Continual Constitutional Convention: How language models revise the documents that govern them
AI Alignment Forum · 13 oct. 2026
The Entangled Dimensions of Decision Theory
LessWrong · 30 iul. 2026
The Double-Edged Sword of Information: Revealed versus Hidden Lotteries in School Choice
arXiv · 30 sept. 2026