Tag
The paper proposes RATTL, a framework that adjusts an agent's caution based on its Bayesian belief uncertainty using Wasserstein distance for safe sequential decision making, applicable to LLM-based systems.
Raven-Agent is the first autonomous trading agent for prediction markets, featuring an explicit belief-to-trade layer. It achieves positive returns on a controlled replay, bridging the gap between calibrated forecasts and profitable trading.
This paper presents a decision-theoretic framework for detecting data leakage in predictive models using only model outputs and outcomes, proving that certain leakage types can be identified without external benchmarks or training code.
This paper argues that aggregating moral evaluations for AI value alignment must account for contextual factors, showing that ignoring context can lead to violations of the weak Pareto principle, analogous to Simpson's paradox.
This paper formalizes the concept of Bayes-sufficient representations in supervised learning, defining when a representation retains exactly the information needed for Bayes-optimal prediction under a given loss function. It introduces the Bayes quotient as a canonical loss-dependent object and connects the framework to property elicitation, illustrating distinctions between sufficiency, minimality, and excess retained information through experiments.
This paper introduces a tree-based formal framework for modeling complementarity in multi-agent human-AI interactions, proving that complementarity is attainable in regression but obstructed in classification under natural conditions on local aggregation and loss functions.
Researchers evaluate 28 LLMs on the St. Petersburg game to distinguish between outcome-level resemblance and mechanism-level alignment in risk decision-making, finding that LLMs often produce human-like bids without underlying human-consistent reasoning mechanisms. The study demonstrates that behavioral alignment can be superficial, urging high-stakes evaluations to go beyond outcome similarity.
This paper solves a COLT open problem by providing an optimal gap-dependent regret algorithm for private stochastic decision-theoretic online learning, achieving the lower bound of order (log K)/Δ_min + (log K)/ε.
This paper presents the first implementation of an infra-Bayesian reinforcement learning agent, demonstrating that it outperforms classical RL in worst-case regret and handles Newcomb's problem optimally, offering a step toward robustness under model misspecification.
This paper derives tight theoretical bounds for human-AI teams, proving when confidence-based aggregation leads to complementarity and establishing impossibility results under specific error correlations.