A new paper finds that LLMs don't sample uniformly from valid solutions when multiple correct answers exist — they have strong defaults driven by the order in which information is presented. Different models share the same defaults. Randomizing presentation increases answer breadth by 26–42 percentage points with only a 3–4 point validity loss.
Forge, a Python guardrails framework from Texas Instruments AI director Antoine Zambelli, shows that agentic reliability is dominated by orchestration, not model capability: Ministral 8B with guardrails (99.3%) outperforms Claude Sonnet without them (87.2%). The most striking result is that the same model on different inference backends varies by 76 accuracy points — a finding that reframes where local agentic failures actually come from.