← back to paper

Review history

arxiv: 2509.04796 · 2 revisions

Knowledge Collapse in LLMs: When Fluency Survives but Facts Fail under Recursive Synthetic Training

1. 2026-08-15 CONDITIONAL MODERATE v1.4.0-alltime-deepseek-medium novelty 5.0
   159350 ms 17043 in 18322 out 2026-08-15T16:27:34.982548+00:00
2. 2026-08-05 CONDITIONAL MODERATE v1.4.0-alltime-deepseek-medium novelty 5.0
   145601 ms 17023 in 15756 out 2026-08-05T05:53:18.181067+00:00