Search papers, labs, and topics across Lattice.
Faculty of Electronics and Information Technology, Warsaw University of Technology, Poland
4
0
5
Selective debate using confidence gating not only boosts performance in argument relation classification but also yields interpretable outputs that traditional methods lack.
KG-CFR achieves over 95% resilience against argument degradation in multi-agent debates, redefining stability in adversarial settings.
Multi-agent debate unlocks significantly better argument classification from LLMs, even without fine-tuning, by surfacing hidden ambiguities that single-agent models miss.
Forget limited scope real-world studies: this LLM-powered agent platform lets you simulate entire social media communities to dissect affective polarization with unprecedented control and granularity.
