cost-aware

GraphDx is a cost-aware, knowledge-enhanced multi-agent framework for sequential diagnosis that uses LLM-constructed medical knowledge graphs and three collaborative agents to improve diagnostic success rates and reduce test costs.

This paper introduces a cost-aware evaluation framework for security AI agents, measuring not just success rate but also inference and tool costs. It finds that offensive CTF performance scales with compute, while defensive SOC tasks depend more on disciplined tool use than raw reasoning budget.

This paper introduces a cost-aware bandit framework for mobile crowdsensing that models worker performance as an increasing-then-converging function and handles unknown costs under budget constraints.

This paper proposes EcoSpec, a cost-aware speculative decoding framework for Mixture-of-Experts models that accounts for expert activation costs during draft selection. It achieves up to 1.62× speedup on large-scale MoE models like DeepSeek-V3.1, Qwen3-235B-A22B, and GPT-OSS-120B by reducing expert footprint without modifying the target model's verification rule.

RouteRec is a framework for evaluating request-level hard selection versus item-level learned aggregation of heterogeneous recommender agents, including an LLM reranker, under cost constraints. Experiments on MovieLens-1M show that item-level aggregation significantly outperforms request-level selection.

Introduces TSRouter, a graph-based dynamic routing framework that selects the optimal modality (LLM or VLM) and model for time-series reasoning tasks, achieving 16-46% relative improvement over baselines and demonstrating zero-shot generalization.

SAGEAgent is an LLM-based clinical agent that sequentially decides which diagnostic modalities to acquire for cancer patients to balance predictive accuracy with clinical invasiveness, reducing acquisition burden by 55% while maintaining competitive survival prediction performance.

AdaStop is a cost-aware early stopping framework for DNN test selection that optimally stops labeling when the marginal fault discovery rate falls below a threshold, achieving 65-84% fault discovery using only 9-31% of the labeling budget.

This paper introduces LearnStop, a lightweight checkpoint stopper for reasoning models that predicts prefix correctness from online features, and finds that learned stopping provides value over scalar rules only when many questions become correct early without a single reliable scalar signal.

Proposes a two-stage cascaded framework for cost-aware LLM serving that clusters queries and routes them to cost-effective models, then escalates low-quality outputs to stronger models. Retains 97-99% of accuracy while reducing inference cost.

Translate-R1 introduces a reinforcement learning approach for cost-aware translation tool use in LLMs, where the model learns to decide when to translate inputs based on its own comprehension and a cost-sensitivity parameter, achieving Pareto-optimal trade-offs across multiple languages.