← back to paper

Review history

arxiv: 2607.14306 · 2 revisions

Tracing LLM Behavior to the Training Data with Empirical Next-Token Distributions

1. 2026-08-04 CONDITIONAL MODERATE v1.3.0-alltime-deepseek novelty 6.0
   163298 ms 13325 in 16829 out 2026-08-04T01:38:43.627819+00:00
2. 2026-08-02 CONDITIONAL MODERATE v1.3.0-alltime-deepseek novelty 6.0
   132274 ms 13717 in 12904 out 2026-08-02T02:28:49.185688+00:00