Founder here. VOR isn’t claiming LLMs don’t hallucinate — it enforces that ungrounded answers never leave the runtime. The model proposes, deterministic gates decide (answer / abstain / conflict), with replayable audits. This is a public demo meant to be challenged; I’m especially interested in failure cases, adversarial packs, or places this would break in real stacks.
Founder here. VOR isn’t claiming LLMs don’t hallucinate — it enforces that ungrounded answers never leave the runtime. The model proposes, deterministic gates decide (answer / abstain / conflict), with replayable audits. This is a public demo meant to be challenged; I’m especially interested in failure cases, adversarial packs, or places this would break in real stacks.