Debate, Deliberate, Decide (D3): A Cost-Aware Adversarial Framework for Reliable and Interpretable LLM Evaluation

Documenti analoghi

• ReDAct: Uncertainty-Aware Deferral for LLM Agents
  di: Piatrashyn, Dzianis, et al.
  Pubblicazione: (2026)
• Context, Reasoning, and Hierarchy: A Cost-Performance Study of Compound LLM Agent Design in an Adversarial POMDP
  di: Bogdanov, Igor, et al.
  Pubblicazione: (2026)
• AutoMedic: An Automated Evaluation Framework for Clinical Conversational Agents with Medical Dataset Grounding
  di: Oh, Gyutaek, et al.
  Pubblicazione: (2025)
• MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety
  di: Wen, Xiaoyu, et al.
  Pubblicazione: (2026)
• GAMBIT: A Three-Mode Benchmark for Adversarial Robustness in Multi-Agent LLM Collectives
  di: Mercier, Alexandre Le, et al.
  Pubblicazione: (2026)