An adaptive budget allocation algorithm for LLM-augmented surveys learns question-level LLM reliability on the fly from human labels and reduces labeling waste from 10-12% to 2-6% compared to uniform allocation.
hub
write newline
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
hub tools
verdicts
UNVERDICTED 11representative citing papers
A Fenchel-Young loss formulation turns data-driven inverse optimization into a differentiable problem solvable by gradient methods, with claimed theoretical guarantees and superior empirical performance on noisy data.
Extends Thompson sampling analysis to Borel MDPs via a three-term regret decomposition and shows exponential convergence of residual regret to zero under extended assumptions.
TimeMark is a trustworthy time watermarking framework that achieves exact generation-time recovery from AI-generated content with theoretically perfect accuracy by using time-dependent cryptographic keys, random non-stored bit sequences, and two-stage encoding with error-correcting codes.
LLM-MAS uses prompt-embedded design choices to drive multi-agent LLM simulations modeled as a controlled Markov chain, with an on-trajectory algorithm for zeroth-order gradient-based optimization of steady-state performance.
Different binarization extended formulations for network flow MIPs cause large differences in solver performance that the authors attribute to cutting-plane generation, with a family of mixed-integer rounding inequalities showing particular benefit.
Derives necessary and sufficient conditions for vanishing regret in the censored data-driven newsvendor under a DRO ambiguity set defined by the max historical order quantity, and proposes a near-optimal adaptive algorithm with finite-sample bounds.
A PPO reinforcement learning method using atomic actions, partially-shared policies, and queueing-informed value approximation scales inpatient overflow optimization to hospital systems with 20 patient classes and wards, matching or beating benchmarks where prior methods fail.
SCOPE is a parameter-free splicing-based algorithm for sparsity-constrained optimization of strongly convex smooth objectives that achieves linear convergence and exact support recovery without relying on RIP-type conditions.
RecPIE jointly optimizes recommendation predictions and LLM-generated natural-language explanations via alternating training and reinforcement learning, yielding 3-4% accuracy gains and higher human preference on Google Maps POI data.
SFLA strengthens an existing convex inner-approximation for RHS-WDRJCC, reducing constraints and tightening ancillary variables to achieve faster computation with no added conservativeness and potential improvement over W-CVaR.
citing papers explorer
-
Adaptive Budget Allocation in LLM-Augmented Surveys
An adaptive budget allocation algorithm for LLM-augmented surveys learns question-level LLM reliability on the fly from human labels and reduces labeling waste from 10-12% to 2-6% compared to uniform allocation.
-
A Fenchel-Young Loss Approach to Data-Driven Inverse Optimization
A Fenchel-Young loss formulation turns data-driven inverse optimization into a differentiable problem solvable by gradient methods, with claimed theoretical guarantees and superior empirical performance on noisy data.
-
Thompson Sampling for Infinite-Horizon Discounted Decision Processes
Extends Thompson sampling analysis to Borel MDPs via a three-term regret decomposition and shows exponential convergence of residual regret to zero under extended assumptions.
-
TimeMark: A Trustworthy Time Watermarking Framework for Exact Generation-Time Recovery from AIGC
TimeMark is a trustworthy time watermarking framework that achieves exact generation-time recovery from AI-generated content with theoretically perfect accuracy by using time-dependent cryptographic keys, random non-stored bit sequences, and two-stage encoding with error-correcting codes.
-
Optimizing Service Operations via LLM-Powered Multi-Agent Simulation
LLM-MAS uses prompt-embedded design choices to drive multi-agent LLM simulations modeled as a controlled Markov chain, with an on-trajectory algorithm for zeroth-order gradient-based optimization of steady-state performance.
-
Cutting Planes for Binarized Network Flow Problems
Different binarization extended formulations for network flow MIPs cause large differences in solver performance that the authors attribute to cutting-plane generation, with a family of mixed-integer rounding inequalities showing particular benefit.
-
The Data-Driven Censored Newsvendor Problem
Derives necessary and sufficient conditions for vanishing regret in the censored data-driven newsvendor under a DRO ambiguity set defined by the max historical order quantity, and proposes a near-optimal adaptive algorithm with finite-sample bounds.
-
Inpatient Overflow Management with Proximal Policy Optimization
A PPO reinforcement learning method using atomic actions, partially-shared policies, and queueing-informed value approximation scales inpatient overflow optimization to hospital systems with 20 patient classes and wards, matching or beating benchmarks where prior methods fail.
-
Sparsity-Constraint Optimization via Splicing Iteration
SCOPE is a parameter-free splicing-based algorithm for sparsity-constrained optimization of strongly convex smooth objectives that achieves linear convergence and exact support recovery without relying on RIP-type conditions.
-
Can Explanations Improve Recommendations? Evidence from Prediction-Informed Explanations
RecPIE jointly optimizes recommendation predictions and LLM-generated natural-language explanations via alternating training and reinforcement learning, yielding 3-4% accuracy gains and higher human preference on Google Maps POI data.
-
Strengthened and Faster Linear Approximation to Joint Chance Constraints with Wasserstein Ambiguity
SFLA strengthens an existing convex inner-approximation for RHS-WDRJCC, reducing constraints and tightening ancillary variables to achieve faster computation with no added conservativeness and potential improvement over W-CVaR.