Research map / Multi-agent systems and games / Games and equilibria
Risk-Sensitive Agent Compositions
Guruprerana Shabadi, Rajeev Alur
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase number of samples from 500 to 10^4 (100 buckets) (ablation) | BucketedVaR · Benchmarks, VaR alpha=0.1 | empirical quantile: stabilizes around 0.91 | Improved |
| increase number of buckets from 5 to 100 (10^4 samples) (ablation) | BucketedVaR · 16-Rooms, VaR alpha=0.1 | empirical quantile | Improved |
| increase number of agents on path from 8 to 8k, k up to 5 (ablation) | BucketedVaR · 16-Rooms, VaR alpha=0.1 | empirical quantile | No clear effect |
| add correlation between agent losses up to rho = 0.5 (ablation) | BucketedVaR · Correlated Gaussian synthetic path | VaR approximation tightness: reasonable up to rho = 0.5 | No clear effect |
| make losses completely correlated (rho = 1) (ablation) | BucketedVaR · Correlated Gaussian synthetic path | VaR approximation tightness | Hurt |
| use BucketedVaR instead of exact baseline | agent graph path selection · All four benchmarks | optimal path found: same optimal path as the baseline | No clear effect |
| use BucketedVaR instead of exact baseline | VaR estimation · All benchmarks, all risk levels | quantile difference: never more than a couple percentage points | No clear effect |
Closest papers by meaning
- Building a Foundational Guardrail for General Agentic Systems via Synthetic Data (ICLR 2026 · similarity 0.85)
- GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory (NeurIPS 2026 · similarity 0.85)
- Automated Composition of Agents: A Knapsack Approach for Agentic Component Selection (NeurIPS 2025 · similarity 0.85)
- Scalable Safe Policy Improvement for Factored Multi-Agent MDPs (ICML 2024 · similarity 0.85)
- Test-time Risk Adaptation with Mixture of Agents (NeurIPS 2026 · similarity 0.84)
- Scalable Constrained Policy Optimization for Safe Multi-agent Reinforcement Learning (NeurIPS 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
