🔥 Excited to share Agent World Model in the Wild (WildAWM)! 🌍
Fully open-source synthesis 🔓: No proprietary models needed! We synthesized 341 high-quality environment–task pairs, addressing 3 challenges in task synthesis:
✅ Solvability: At least one agent can successfully complete each task.
🔥 Difficulty: Challenging even for frontier models, with GPT-6-Astra achieving only ~72.7 pass@8.
🎯 Verifier reliability: 95% agreement between task verifiers and agent-as-a-judge ensemble.
📈 Effective RL training: Trained exclusively in these synthetic worlds, Qwen3.6-35B-A3B achieves an average +10.58-point improvement across 8 OOD benchmarks!
Moderna and @Merck today announced positive topline results from the Phase 3 INTerpath-001 trial evaluating adjuvant treatment with intismeran autogene, a novel investigational mRNA-based individualized neoantigen therapy jointly developed by Merck and Moderna, in combination with KEYTRUDA® (pembrolizumab), Merck's anti-PD-1 therapy, in patients with completely resected Stage IIB-IV melanoma.
Read more:
For my last summer’s internship project, we present SEMA (ICLR 2026), a simple and scalable framework for learning highly effective red-teamers to stress-test the security of LLMs and AI agents.
After a long wait, we can finally make our code public! 🎉
https://t.co/I4B6R2URzs
MatrAIx team co-organizers @YuexingHao and @XiaominLi98 delivered a great talk at @Stanford yesterday. First talk after the community released the MatrAIx-8.3B paper!
With our amazing @MatrAIx2026 team swag 🕶️
Thanks for highlighting our work!
The core idea of FlowRL is to shift from reward maximizing to distribution matching via flow balance — encouraging diverse and generalizable reasoning in LLM reasoning.