AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

UniAPO: Unified Multimodal Automated Prompt Optimization

Qipeng Zhu, Yanzhe Chen, Huasong Zhong, Jie Chen, Yan Li, Zhixin Zhang, Junping Zhang, Zhenheng Yang

AAAI 2026 · Large language models › Language, safety and interpretability · approach: In-context learning analysis · 0 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
use UniAPO instead of ERM* for prompt optimizationGPT-4o predictor · LIARF1: 65.2 -> 78.7Improved
use UniAPO instead of ERM* on WebNLGGPT-4o predictor · WebNLGROUGE-L: 52.1 -> 53.2Improved
use UniAPO instead of ERM* on ETHOS (GPT-4o)GPT-4o predictor · ETHOSF1: 95.6 -> 98.1Improved
use UniAPO instead of ERM* with QwenVL2.5-72B predictorQwenVL2.5-72B · LIARF1: 67.4 -> 73.1Improved
use UniAPO instead of ERM* on BBH-navigate with Qwen predictorQwenVL2.5-72B · BBH-navigateF1: 93.3 -> 95.8Improved
add both E-step and M-step to simple-prompt baseline (ablation)UniAPO · Video KE BeautyF1: 36.7 -> 75.2Improved
add either E-step or M-step alone (ablation)UniAPO · Video KE BeautyF1: 36.7 -> 66.3 / 67.8Improved
use UniAPO feedback generator instead of ERM*'s (ablation)UniAPO with ERM* prompt optimizer · Video KE BeautyF1: 68.3 -> 73.1Improved
use UniAPO prompt optimizer instead of ERM*'s (ablation)ERM* feedback generator · Video KE BeautyF1: 68.3 -> 70.7Improved
add long-term feedback memory (ablation)UniAPO · Video KE BeautyF1: 70.9 -> 74.7Improved
add long-term prompt memory (ablation)UniAPO · Video KE BeautyF1: 71.3 -> 74.7Improved
add historical prompts and beam size beyond ERM* (ablation)UniAPO · Video KE (Figure 5b)F1Improved

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