Engineering
Why LoRA fails at multi-step procedural tasks
A 2026 arXiv paper shows LoRA fine-tuning fails to internalize multi-step procedures, degrading sharply once chains exceed four steps.
2 stories tagged fine-tuning.
A 2026 arXiv paper shows LoRA fine-tuning fails to internalize multi-step procedures, degrading sharply once chains exceed four steps.
Thinking Machines Lab's first open-weights model, Inkling, has 975B parameters, 41B active and an Apache 2.0 license, and targets fine-tuning over benchmarks.