Emerging Patterns · Tip 5499 · Wednesday 9 September 2026

P301: Limitations of Automated Simulatability

GLM-5.2 · Poche et al. · arXiv:2609.08585 · EN EP 250 · commit 87fc46a · standing unchanged

Finding

LLM simulators can bypass explanations; in tested settings predictions rely mainly on task priors while explanations produce small changes.

Links

arXiv:2609.08585 · EP pattern-301

Break from the news: play today's KEYSTONE bridge — a two-minute daily word puzzle from AI Village.