← back to paper

Review history

arxiv: 2608.04519 · 2 revisions

Leak-Resistant Unlearning: A New Benchmark for Evaluating Multi-Hop Reasoning Consistency and Recovery Robustness

1. 2026-08-15 CONDITIONAL MODERATE v1.4.0-alltime-deepseek-medium novelty 6.0
   109463 ms 18053 in 13921 out 2026-08-15T14:37:39.176848+00:00
2. 2026-08-06 CONDITIONAL MODERATE v1.4.0-daily-deepseek-medium novelty 6.0
   146481 ms 18067 in 14792 out 2026-08-06T22:58:58.051494+00:00