CVaR-RMD: CVaR Selection Pressure for Robust Metaheuristic Discovery
We propose CVaR-RMD, a risk-sensitive evolutionary framework that uses an LLM to mutate executable heuristic programs and ranks candidates by empirical conditional value-at-risk over heterogeneous evaluation conditions.
ProofPaper ↗
Key points
- Large language models have recently been used to discover heuristics through iterative program search, offering a promising approach to automated algorithm design.
- However, candidates selected by average performance may degrade across instance families, random seeds, and problem scales because mean-based objectives can obscure rare but severe failures.
- Across three combinatorial optimization domains, CVaR-based selection improves upper-tail performance and reduces failure rates while maintaining competitive mean performance.
- These gains persist on larger unseen instances and independently curated within-domain benchmarks
Sources (1)
- [1]CVaR-RMD: CVaR Selection Pressure for Robust Metaheuristic DiscoveryarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 02:43 AM
We propose CVaR-RMD, a risk-sensitive evolutionary framework that uses an LLM to mutate executable heuristic programs and ranks candidates by empirical conditional value-at-risk over heterogeneous evaluation conditions.
Large language models have recently been used to discover heuristics through iterative program search, offering a promising approach to automated algorithm design.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 18, 2026Outcome-Anchored Credit Reallocation for Agentic Knowledge Graph Question Answering
- Oct 18, 2026PRECOCIAL: Pretraining with Code Curriculums via Iterative Search for Accelerated Learning
- Oct 17, 2026A Cross-Domain Evaluation of Prompting Techniques for Large Language Models
- Oct 17, 2026HybridResearcher: Bridging the Structured-Unstructured Boundary for Deep Research Agents
- Oct 16, 2026Characterising and Analysing Retrieval-Augmented Generation Techniques
- Oct 16, 2026Before Retrieval Begins: Safety and Tool-Use Decisions in LLM Agents