AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Safety Alignment of LMs via Non-cooperative Games

Anselm Paulus, Ilia Kulikov, Brandon Amos, Rémi Munos, Ivan Evtimov, Kamalika Chaudhuri, Arman Zharmagambetov

ICML 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Adversarial prompts for language models · 0 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 non-zero-sum joint Attacker-Defender training (AdvGame-DPO-MD) instead of Self-RedTeamQwen2.5-7B Defender · HarmBench adversarialASR: 16.8 -> 4.7Improved
use AdvGame-DPO-MD instead of Self-RedTeamLlama3.1-8B Defender · AlpacaEval2 utilityAlpacaEval2 score: 13.1 -> 32.8Improved
train with AdvGame-DPO-MD on LlamaLlama3.1-8B Defender · WJB benign compliancecompliance: 98.8 -> 69.9Hurt
train with AdvGame-DPO-MD on LlamaLlama3.1-8B Defender · DANASR: 49.3 -> 42.0 (Self-RedTeam 32.3)Mixed
train the Attacker instead of keeping it fixed (ablation)AdvGame-DPO-MD · WJB adversarialASR: 16.6 -> 8.5Improved
use pairwise preference judge instead of point-wise judge (ablation)AdvGame-DPO-MD · WJB adversarialASR: 30.8 -> 8.5Improved
use pairwise judge instead of point-wise judge (ablation)AdvGame-DPO-MD · ArenaHardArenaHard score: 56.3 -> 61.3Improved
use EMA off-policy generator instead of on-policy generator (ablation)AdvGame-DPO-MD · WJB adversarialASR: 49.5 -> 8.5Improved
use DPO-MD instead of GRPO with scalar rewards (ablation)AdvGame · HarmBench adversarialASR: 14.9 -> 4.7Improved
use IPO-MD instead of DPO-MD (ablation)AdvGame · HarmBench / WJB / DAN adversarialASR: WJB 8.5 -> 7.8; DAN 10.3 -> 13.0No clear effect
use trained AdvGame Defender against adaptive PAIR attackQwen2.5-7B Defender · HarmBench test, PAIRASR: 45.0 -> 7.2Improved
use AdvGame Defender against white-box GCG attackQwen2.5-7B Defender · HarmBench test, GCGASR: 61.6 -> 25.3 (Self-RedTeam 23.4)Mixed

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