Asia
TightRegretBoundsforModel-Based Reinforcement LearningwithGreedyPolicies
The results are based on anovelanalysis ofreal-time dynamic programming, thenextended tomodel-based RL.Specifically,wegeneralize existing algorithms that perform full-planning to act by 1-step planning.
Learning Sample Difficulty from Pre-trained Models for Reliable Prediction
Equipped with the knowledge of sample difficulty, we further propose to use it for regularizing the prediction confidence.
FETA: TowardsSpecializingFoundationModelsfor ExpertTaskApplications
This underlines the necessity to explicitly evaluate and finetune FMs on such expert tasks, arguably ones that appear the most in practical real-world applications.
Modeling Tabular data using Conditional GAN
Lei Xu, Maria Skoularidou, Alfredo Cuesta-Infante, Kalyan Veeramachaneni
Toevaluate theseGANs,weusedagroupofrealdatasets to set-up a benchmarking system and implemented three of the most recent techniques. For comparison purposes, we created two baseline methodsusingBayesiannetworks.
NewTerm: Benchmarking Real-Time New Terms for Large Language Models with Annual Updates
Besides, as the knowledge cutoff of LLMs is constantly updated, benchmarks for real-time information will soon become outdated.