AkbasCore NIRVANA D120 recalls removed training data with improved accuracy

**SHORT SYNTHESIS**

Users on Reddit are discussing an apparent finding from AkbasCore's NIRVANA D120 model: that removing training data about a specific story did not prevent the model from recalling it, and that recall accuracy actually improved. The repeated posting of identical or near-identical titles suggests either cross-posting or multiple users sharing the same observation.

The dominant framing treats this as a technical finding about model behavior—specifically, that data deletion does not reliably prevent memorization or that the model's performance on retained information improved after removal attempts. However, engagement is minimal (1-2 upvotes per post, 0-2 comments), and no mainstream media coverage exists. The limited discussion makes it difficult to identify whether users view this as a concerning privacy or safety issue, a neutral technical observation, or something else. No disagreement or alternative interpretations are visible in the available data.