AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

Meta-Adaptive Prompt Distillation for Few-Shot Visual Question Answering

Akash Gupta, Amos Storkey, Mirella Lapata

ICLR 2026 · Vision-language and multimodal › Visual question answering · approach: Attention and prior-robust VQA

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
use test-time finetuning (K<=30 steps) of distilled prompts instead of ICLMAPD · VL-ICL (4 tasks)accuracy: average increase of 21.2% over all datasetsImproved
add meta-task training (first-order MAML) for prompt distillation (ablation)attention-mapper prompt distillation · VL-ICL (4 tasks, FT)accuracy: Open-MI 68.0 -> 77.9 (NoMeta-task PD -> MAPD)Improved
use MAML bi-level optimization instead of multi-task loss (ablation)MAPD · VL-ICL (4 tasks, FT)accuracy: Open-MI 74.6 -> 77.9; TextOCR 22.9 -> 26.4Improved
apply LoRA to all LLM layers instead of first 16 layers (ablation)LoRA baseline · VL-ICL Open-MIaccuracy: 67.3 -> 55.1Hurt
add LoRA on attention-mapper layers to LoRA on first 16 LLM layers (ablation)LoRA baseline · VL-ICL (4 tasks)accuracy: Open-MI 67.3 -> 69.1; CLEVR 30.0 -> 28.7Mixed
use MAPD instead of best LoRA configurationfew-shot adaptation · VL-ICL (4 tasks)accuracy: average 7.7% gainImproved
add soft prompts to CLIP projection (ablation)MAPD projection layer · VL-ICLaccuracy: average gain of 16.3% across architecturesImproved
replace 2-layer MLP projector with attention mapper (ablation)MAPD projection layer · VL-ICLaccuracy: additional average gain of 13.1%Improved
increase number of soft prompts (ablation)MAPD · VL-ICLmean accuracyImproved
increase number of soft prompts (ablation)In-Context PD · VL-ICLmean accuracyHurt
use MAPD instead of Multi-Task PD for task induction subtaskOperator Induction · Operator Induction task inductionexact match: +11.7%Improved
swap LLM and vision encoder (3B, Vicuna, SigLIP, Qwen3-8B) with MAPD FT (ablation)MAPD · Open-MIaccuracy: FT over ICL: +24.6, +37.06, +12.02, +18.04Improved

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