AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

GateRA: Token-aware Modulation for Parameter-Efficient Fine-tuning

Jie Ou, Shuaihong Jiang, Yingjun Du, Cees G. M. Snoek

AAAI 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning · 1 citations (OpenAlex)

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 token-aware gating g(x) to HiRA (ablation)HiRA · Commonsense reasoning (8 tasks, average)accuracy: 86.72 -> 87.53Improved
make the gate input-dependent instead of a static learnable tensor (ablation)GateRA · Commonsense reasoning (8 tasks, average)accuracy: 86.97 -> 87.53Improved
add entropy-based regularization on gate values (ablation)GateRA · Commonsense reasoning (8 tasks, average)accuracy: 87.08 -> 87.53Improved
halve adaptation rank from 32 to 16 (ablation)GateRA · Commonsense reasoning (8 tasks, average)accuracy: 87.53 -> 86.70No clear effect
apply GateRA to both FC and QKV rather than FC only (ablation)GateRA · Commonsense reasoning (8 tasks, average)accuracy: 86.53 -> 87.53Improved
apply GateRA to QKV rather than K, Q or V alone (ablation)GateRA · Commonsense reasoning (8 tasks, average)accuracy: 83.88 -> 86.50Improved
replace HiRA with GateRALLaMA-2-7B · Commonsense reasoning (8 tasks, average)accuracy: 81.42 -> 82.52Improved
replace HiRA with GateRALLaMA-3-8B · GSM8Kaccuracy: 70.81 -> 72.11Improved
replace HiRA with GateRALLaMA-3-8B · ConvAI2 (average of 6 metrics)average score: 47.80 -> 48.61Improved
replace HiRA with GateRALLaMA-3-8B · PIQAaccuracy: 89.70 -> 89.45No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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