Posts tagged #lora

• Adversarial Passes That Killed Claims

  Apr 22, 2026

  Two hypotheses that started as clean ideas and ended as documented failures: the DPO-CLM orthogonal-complement hypothesis, and the cross-probe srank claim retracted as a length-bias artifact.

  #ml-research #lora #dpo #falsification #methodology #negative-results

• Stable Rank as an Overfitting Signature in LoRA Fine-Tuning

  Apr 18, 2026

  Why we picked stable rank to detect overfitting geometry in DPO vs CLM fine-tuning, how it connects to "alignment geometry," and what the BitFit baseline was there to check.

  #ml-research #lora #fine-tuning #dpo #geometry #stable-rank