Increasing LoRA rank (8, 16 → 64) didn’t improve results — why?

arXiv:2604.27796v1 Announce Type: new Abstract: Exponential growth in the scale of modern foundation models has led to the widespread adoption of Low-Rank Adaptation (LoRA) as a

<table> <tr><td> <a href="https://www.reddit.com/r/LocalLLaMA/comments/1s59rct/rys_part_3_llms_think_in_geometry_not_language/"> <img alt="RYS Part 3: LLMs think in geometry, not

<!-- SC_OFF --><div class="md"><p>I have a question I hope some of you llm experts can enlighten me on.</p> <p>In my baby understanding of LLMs there are a bunch of linear layers

<!-- SC_OFF --><div class="md"><p>That foodtruck bench post showing deepseek v4 matching gpt-5.2 at 17x cheaper got me thinking. if frontier cloud models are that overpriced for

<!-- SC_OFF --><div class="md"><p>Been losing my mind trying to figure out why my H200 fine-tunes look fine during training then perform terribly. Validation loss looks

arXiv:2606.12883v1 Announce Type: new Abstract: In Low-Rank Adaptation (LoRA), the scaling factor $\alpha$ is often treated as a mere complement to the learning rate, yet its role

arXiv:2609.32002v1 Announce Type: new Abstract: The rank $r$ in LoRA is widely treated as a capacity control: a smaller rank is assumed to yield a simpler model that generalizes

<!-- SC_OFF --><div class="md"><blockquote> <p>Wait, I need to be careful about the &quot;no_think&quot; tag in the system prompt. The system prompt says <code>/no_think</code>.