This paper proposes two strategies to improve feedback efficiency of reinforcement learning from human feedback (RLHF) in diffusion models.

The paper introduces AsyncTool, a new benchmark designed to evaluate LLM agents' ability to handle multiple, concurrent tasks with delayed tool feedback, demonstrating that asynchronous coordination i…

This paper surveys 1,250 arXiv papers on self-improving AI systems, categorizing them based on what they improve and the degree of loop closure. It identifies a distinctive feature of self-evaluation…

This paper proposes RLAES, a unified language model framework using reinforcement learning for essay scoring and feedback generation, with methods including Rubric-based Feedback Evaluation (RFE), Ada…

This paper develops a convergence theory and runtime bound for closed-loop generative selection in computational drug discovery, showing that elitism makes the search absorbing and proving almost-sure…

The paper proposes a coherent inverse reinforcement learning (IRL) method to improve large behavior models for robotic control, achieving superior sample efficiency and performance on complex sparse m…

The paper introduces an Item Response Theory (IRT)-based indicator that effectively identifies likely mislabeled items in existing LLM benchmarks, revealing systematic errors in labeling and model spe…

This paper analyzes multi-model self-consuming training, showing that while human curation helps individual models, cross-model interactions can degrade long-term alignment by dampening or inverting t…

This paper improves an algorithm for computing lower bounds on the margin of a Single Transferable Vote (STV) election, making risk-limiting audits more practical.

The paper introduces FormInv, a measurement protocol that reveals significant semantic inconsistencies in existing mathematical reasoning benchmarks, showing that standard accuracy metrics fail to cap…

This paper investigates the necessity of interaction for order-optimal 1-bit mean estimation in nonparametric finite-moment classes.

The paper shows that deterministic cache eviction cannot ensure consistent serving-time error estimation and proposes a randomized approach to restore identifiability and provide error certificates.

This paper analyzes tool-calling in LLM agents, demonstrating that evaluation results are highly sensitive to implementation details and proposing new techniques to significantly improve the efficienc…

The paper introduces a Jacobian-based spectral audit to evaluate neural operators, demonstrating that standard prediction error metrics fail to capture crucial local dynamical structures and operator…

This paper measures the impact of procedural skills on LLM agents, distinguishing between improvements and regressions, and identifies causes of regression.

The paper critiques current AI benchmarking practices for low-resource settings, arguing that evaluation must shift focus from isolated model performance to the holistic performance of the deployed sy…

This paper analyzes the use of generative AI for feedback in higher education based on 2988 instances from Estonian bachelor students, finding that students generally find it helpful but not a self-co…