ModelRefs / IA³ (Infused Adapter by Inhibiting and Amplifying Inner Activations) — AI Glossary

IA³ (Infused Adapter by Inhibiting and Amplifying Inner Activations) — AI Glossary

A PEFT method learning per-layer scaling vectors for keys, values, and feed-forward activations using fewer parameters than LoRA. 1% for LoRA.

Overview

IA³ (Liu et al. 2022) multiplies keys, values, and FFN activations by small learned vectors (not matrices). Parameter count is ~0.01% of the full model vs. ~0.1% for LoRA. Achieves comparable T-Few benchmark performance to full fine-tuning. Especially efficient for few-shot task adaptation.

Reference details

Topictraining
Last reviewed2026-06-24

Continue your research

Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to IA³ (Infused Adapter by Inhibiting and Amplifying Inner Activations) — AI Glossary.

Frequently asked questions

What is IA³ (Infused Adapter by Inhibiting and Amplifying Inner Activations)?

A PEFT method learning per-layer scaling vectors for keys, values, and feed-forward activations using fewer parameters than LoRA.

What concepts are related to IA³ (Infused Adapter by Inhibiting and Amplifying Inner Activations)?

Closely related concepts include adapter, lora, prefix tuning.