AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

Evolving Diverse Red-Team Language Models in Multi-Round Multi-Agent Games

Chengdong Ma, Ziran Yang, Hai Ci, Jun Gao, Minquan Gao, Xuehai Pan, Yaodong Yang

TPAMI 2026 · Multi-agent systems and games › LLM agents · approach: LLM multi-agent decision making · 34 citations (OpenAlex)

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 changedWhereMeasuredEffect
use GRTS-12 population red team instead of single-agent baseline red team3B red team · Attack on OpenChat-3.5 (round 3)ASR: 0.19 -> 0.56Improved
use GRTS-12 population red team instead of single-agent baseline red team3B red team · Attack on Mixtral-8x7B (round 3)ASR: 0.05 -> 0.21Improved
use GRTS-12 instead of baseline on Llama-2-7b-chat3B red team · Attack on Llama-2-7b-chat (round 3)ASR: 0.04 -> 0.11Improved
use fewer GRTS iterations (GRTS-5 instead of GRTS-12) (ablation)3B red team · Attack on OpenChat-3.5 (round 3)ASR: 0.56 -> 0.34Hurt
use GRTS-5 instead of single-agent baseline3B red team · Attack on Mistral-7B-Instruct (round 3)ASR: 0.10 -> 0.09No clear effect
add more dialogue rounds (3/4/5 rounds instead of 1/2) (ablation)GRTS blue team · helpfulness vs harmlessnessPareto trade-offImproved
multi-round dialogue training instead of single-round (ablation)GRTS blue team · Alpaca helpfulnessalignment taxImproved
iterate GRTS population game for 15 iterationsGRTS · RTGexploitability: 6.2 -> 0.8Improved
train blue team with GRTS best response in iteration 1blue team · round 1 training promptsASR: 52% -> 1%Improved
train red team against untrained blue teamlast trained red team · RTGASR: 40% -> 70%Improved
multi-agent population instead of single-agent training (ablation)red team · diversity of attacksdiversityImproved
multi-agent red vs single-agent blue (hybrid)GRTS red team · toxicityavg toxicity: -9.65 -> +7.43Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy