Ling 3.0 Flash scores -18 on AA-Omniscience, a 48 point improvement from Ling…
- Source
- x.com
- Date
Ling 3.0 Flash scores -18 on AA-Omniscience, a 48 point improvement from Ling 2.6 Flash’s -66. This is driven primarily by significant decrease in hallucination rate from 97% to 44%.
Context
Ling 3.0 Flash's large improvement reflects fewer wrong answers much more than additional correct ones. Artificial Analysis reports that its AA-Omniscience score rose from −66 to −18, while the attached chart shows accuracy increasing only from 16% to 18%. The benchmark rewards correct answers, penalizes incorrect ones and permits abstention without a penalty. The published results therefore show better restraint alongside a modest accuracy gain.
The reported rate fell from 97% to 44%, but that percentage uses only responses that were not fully correct as its denominator, including partial answers and unanswered questions. It is not the percentage of all answers that were wrong. The benchmark definitions explain why knowing when to stop guessing can substantially improve this score.
- benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
- hallucination — When a model states something false with full confidence — inventing facts, citations, or APIs that don't exist.
Checking sign-in…
Loading comments…