AI Paper Atlas

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

The Visual Prism: Refracting Images into Parallel Multilingual Descriptions with Structured Visual Guidance

Chengpeng Fu, Xiaocheng Feng, Yichong Huang, Wenshuai Huo, Baohang Li, Yang Xiang, Ting Liu

AAAI 2026 · Large language models › Instruction and fine-tuning · approach: Calibration and knowledge in LLM tuning

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
fine-tune on PRISMS pseudo-parallel corpus instead of zero-shotLLaMA3-8B · EN-FR Flores200BLEU: 31.19 -> 37.06Improved
fine-tune on PRISMS pseudo-parallel corpus instead of human parallel corpusLLaMA3-8B · EN-FR average over 4 test setsBLEU: 31.56 -> 30.28Hurt
fine-tune on PRISMS pseudo-parallel corpus instead of human parallel corpusLLaMA3-8B · FR-EN Flores200BLEU: 41.32 -> 41.41No clear effect
use vanilla MLLM captions generated independently per language (ORI) as parallel corpus (ablation)LLaMA3-8B · EN-FR Flores200BLEU: 31.19 -> 22.97Hurt
add structured visual guidance prompt to a vanilla MLLM without adherence training (ablation)caption generation · EN-FR Flores200BLEU: 22.97 -> 2.34Hurt
add QE filtering to SVG-only descriptions (ablation)caption generation · EN-FR Flores200BLEU: 2.34 -> 24.24Improved
add SVG adherence training (pseudo-data SFT) on top of SVG prompting (ablation)caption generation · EN-FR Flores200BLEU: 24.24 -> 37.06Improved
replace pseudo-data SFT with Iter-DPO for SVG adherence (ablation)caption generation · EN-ZH Flores200BLEU: 41.80 -> 41.71No clear effect
replace pseudo-data SFT with Iter-DPO for SVG adherence (ablation)caption generation · ZH-EN Flores200BLEU: 23.42 -> 23.90No clear effect
add post-processing filter that drops descriptions with inconsistent triplet counts (ablation)PRISMS corpus · EN-FR Flores200BLEU: 35.01 -> 37.06Improved
use PRISMS pseudo-parallel data to initialize unsupervised MTXLM UNMT · UNMT Flickr2017BLEU: 38.57 -> 41.58Improved
fine-tune on PRISMS corpusLLaMA3-8B · FR-EN Flores200BLEU: 37.39 -> 41.41Improved

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