DoRA: which component does LoRA actually adapt?

MCQ·Hard·4.0 · 0·~1 min·Relevant atCohere Databricks Meta Microsoft

Attempt it

Quick reference

TL;DR

DoRA splits each weight into magnitude and direction, learns the magnitude directly, and applies LoRA only to the direction. This closes more of the gap to full fine-tuning.

Memory aid

Easy to grasp

Picture aiming a flashlight. A weight has two parts: how far the beam reaches, its length or magnitude, and where you point it, its direction. Plain LoRA tries to learn both at once with a small budget, so it gets pulled two ways and underfits. DoRA hands the length its own simple dial, a single number you turn directly, and spends the small LoRA budget purely on steering the beam. Because length and aim no longer fight over the same budget, the steering comes out cleaner. The result lands closer to retraining the whole flashlight from scratch, while still touching almost no extra parameters.

Key concepts

DoRA weight decomposition LoRA adapters Parameter-efficient fine-tuning

Detailed answer & concept explanation

~7 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Quick scan · 2–4 min

Everything important, quickly.

4 min: state the decomposition, identify which component LoRA touches, explain the magnitude-direction motivation, then distinguish DoRA from AdaLoRA and QLoRA distractors.

Side-by-side

Method	What it parameterises	Key idea
LoRA	Low-rank update BA on the full weight	Couples magnitude and direction in one low-rank delta
DoRA	Direct magnitude scalar plus LoRA on the direction	Decouples length from aim to mimic full fine-tuning
AdaLoRA	SVD-form update with adaptive rank per layer	Allocates rank budget by importance via singular values
QLoRA	4-bit frozen base plus bf16 low-rank factors	Memory trick, orthogonal to the decomposition idea

Real-world examples

Real products, models, and research that use this idea.

Hugging Face PEFT ships DoRA as a flag (use_dora=True) on its LoRA config, so teams fine-tuning Llama 4 can A/B it against plain LoRA with one line.
Unsloth and Axolotl both expose DoRA toggles in their fine-tuning recipes for open models like Mistral and Qwen variants.
NVIDIA NeMo and the broader 2026 PEFT tooling list DoRA among recommended adapters for instruction-tuning open-weight bases such as Llama 4.
Practitioners commonly benchmark DoRA against LoRA on commonsense-reasoning suites, where DoRA's magnitude-direction split shows a few-point accuracy lift at matched parameters.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QWhy does decoupling magnitude from direction help LoRA approach full fine-tuning?

QHow does DoRA compose with QLoRA, and is stacking them sensible?

QWhat overhead does DoRA add at training time and can you merge it for serving?

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Confusing DoRA with AdaLoRA's SVD update or QLoRA's 4-bit base. DoRA's contribution is structural, a magnitude versus direction split, not a quantisation or rank allocation trick.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

References

Primary sources. Skim if you want the original framing.