AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

IA2: Alignment with ICL Activations improves Supervised Fine-Tuning

Aayush Mishra, Daniel Khashabi, Anqi Liu

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Add IA2 activation-alignment priming before SFT (ablation)SFT with LoRA · single-token tasks, N=4accuracy and ECE: SST2 acc 65.2 -> 90.4; ECE 0.22 -> 0.06Improved
Add IA2 priming before SFT on multi-token tasks (ablation)SFT · GSM8K, GSM8Ks, MATH Algebra, SciQ, N=4accuracy: GSM8K 70.9 -> 73.6; HMathA 50.4 -> 55.3Improved
Add IA2 priming before SFT versus using ICL at inferenceQwen3-4B-Base math · GSM8K, HMathAaccuracy: GSM8K 76.4 -> 73.6 (ICL vs IA2->SFT); HMathA 60.4 -> 55.3Hurt
Train with IA2 only (no SFT) (ablation)Qwen3-4B-Base · SciQ generation, N=4accuracy: 40.8 -> 06.9Hurt
Add IA2 loss jointly with SFT on ICL responses (ablation)SFT on ICL responses · GSM8K, N=4accuracy: 66.4 -> 77.0Improved
Use soft-label KD on ICL responses instead of SFT only (ablation)Llama SFT · GSM8K and SST2accuracy: GSM8K 28.2 -> 36.1; SST2 52.3 -> 55.9Mixed
Train only QK matrices instead of QKO in LoRA (ablation)IA2 -> SFT, Llama-1B · FinS, PoemS, SST2accuracy: FinS 89.8 -> 50.4Hurt
Increase LoRA rank from 8 to 16 (ablation)IA2 -> SFT · GSM8K, SST2accuracy: GSM8K 73.6 -> 74.1No clear effect
Swap LoRA target from IA2 to (IA)^3 adaptersIA2 -> SFT · SciQr/QASCraccuracy: QASCr 32.8 -> 48.6 vs SFT onlyImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy