We present a submission to the SemEval 2025 shared task on unlearning sensitive content from LLMs. Our approach employs negative preference optimization using low-rank adaptation. We show that we can utilize this combination to efficiently compute ad...
Large Language Models (LLMs) have catalyzed vibe coding, where users leverage LLMs to generate and iteratively refine code through natural language interactions until it passes their vibe check. Vibe check is tied to real-world human preference and g...
Positions of chess players in intransitive (rock-paper-scissors) relations are considered. Namely, position A of White is preferable (it should be chosen if choice is possible) to position B of Black, position B of Black is preferable to position C o...
Long-form question answering (LFQA) demands nuanced evaluation of multi-sentence explanatory responses, yet existing metrics often fail to reflect human judgment. We present LFQA-HP-1M, a large-scale dataset comprising 1.3M human pairwise preference...
Direct preference optimization (DPO) is widely used as a simple and stable method for aligning large language models (LLMs) with human preferences. This paper investigates a generalized DPO loss that enables a policy model to match the target policy...
Parce que généralement on dit qu’une personne vas forcément trouver chaussure à son pied même si elle a des préférences comme une personne petite, mince, enveloppé ou grande ...
"Ma préférence" is a song by French singer-songwriter Julien Clerc. It was released as the first single from his 1978 album Jaloux. The song is composed
européennes en Belgique: géographie des votes de préférence [The European Elections in Belgium: Geography of the Preference Votes] (Report) (in French). Courrier
We present the effect of adapting to human preferences on trust in a human-robot teaming task. The team performs a task in which the robot acts as an action recommender to the human. It is assumed that the behavior of the human and the robot is based...
Modeling user sequential behaviors has recently attracted increasing attention in the recommendation domain. Existing methods mostly assume coherent preference in the same sequence. However, user personalities are volatile and easily changed, and the...
It remains an open question how to determine the winner of an election when voter preferences are incomplete or uncertain. One option is to assume some probability space over the voting profile and select the Most Probable Winner (MPW) -- the candida...
Trae Young has the Washington Wizards as his top preferred trade destination, sources tell ESPN. Atlanta is working to trade Young, who has a $49 million player option in offseason that would make him...
Preferential attachment (PA) models of network structure are widely used due to their explanatory power and conceptual simplicity. PA models are able to account for the scale-free degree distributions observed in many real-world large networks throug...
We introduce a family of one-dimensional geometric growth models, constructed iteratively by locally optimizing the tradeoffs between two competing metrics, and show that this family is equivalent to a family of preferential attachment random graph...
A variation of the preferential attachment random graph model of Barabási and Albert is defined that incorporates planted communities. The graph is built progressively, with new vertices attaching to the existing ones one-by-one. At every step, the...
We present a revealed preference characterization of marital stability where some couples are committed. A couple is committed if they can divorce only with mutual consent. We provide theoretical insights into the potential of the characterization fo...
Sep 1, 2010 · file:/ is invalid as far as I know file:// is the prefix for the file protocol file:/// is the prefix for the file protocol, plus a leading / pointing to the root directory of the current drive (On Windows) or …
Hirata may refer to: Hirata, Fukushima, village in Fukushima Prefecture, Japan Hirata, Gifu, former town in Gifu Prefecture, Japan Hirata, Shimane, former
The success of the reward model in distinguishing between responses with subtle safety differences depends critically on the high-quality preference dataset, which should capture the fine-grained nuances of harmful and harmless responses. This motiva...
We provide an axiomatic characterization of lexicographic preferences over the set of all random availability functions using two assumptions. The first assumption is strong monotonicity, which in our framework is equivalent to the strong dominance p...