← back to paper

Review history

arxiv: 2607.09999 · 2 revisions

Silent Failures in Quantized LLM Reasoning: A Taxonomy-Based Analysis of Hollow Convergence and Failure Mode Shifts

1. 2026-08-02 REJECT MODERATE v1.3.0-alltime-deepseek novelty 6.0
   221493 ms 10354 in 21608 out 2026-08-02T07:25:51.994457+00:00
2. 2026-07-14 CONDITIONAL MODERATE v1.1.0-grok45 novelty 6.0
   50037 ms 16047 in 4322 out 2026-07-14T01:08:30.536816+00:00