AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

CLIP-DPO: Vision-Language Models as a Source of Preference for Fixing Hallucinations in LVLMs

Yassine Ouali, Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos

ECCV 2024 · Vision-language and multimodal › Multimodal LLMs · approach: Preference tuning against hallucination · 24 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 CLIP-DPO fine-tuningLLaVA-1.5 7B · AMBERAMBER score: 83.5 -> 89.6Improved
add CLIP-DPO fine-tuningMobileVLM-v2 1.7B · AMBERAMBER score: 81.4 -> 85.3Improved
use CLIP-DPO instead of HA-DPOLLaVA-1.5 7B · AMBERAMBER score: 86.7 -> 89.6Improved
add CLIP-DPO fine-tuningMobileVLM-v2 1.7B · AMBER generativeCHAIR: 3.8 -> 4.2Hurt
add CLIP-DPO fine-tuningLLaVA-1.5 7B · Zero-shot classification (9 datasets avg)top-1 accuracy: 40.0 -> 47.4Improved
add HA-DPOLLaVA-1.5 7B · Zero-shot classification (9 datasets avg)top-1 accuracy: 40.0 -> 39.7No clear effect
add CLIP-DPO fine-tuningLLaVA-1.5 7B · LLaVA benchmark (avg of 6)average score: 68.8 -> 67.9Hurt
add CLIP-DPO fine-tuningMobileVLM-v2 7B · LLaVA benchmark (avg of 6)average score: 72.1 -> 72.3No clear effect
use alternative DPO-style loss (IPO, KTO, SLIC, cDPO) instead of DPO beta=0.1 (ablation)CLIP-DPO · AMBER (MobileVLM-v2 1.7B)AMBER score: 85.3 -> 83.1 to 83.6No clear effect
use a stronger CLIP scorer (ViT-H/14 DFN instead of ViT-L/14) (ablation)CLIP-DPO · AMBER (MobileVLM-v2 1.7B)AMBER score: 79.8 -> 85.3Improved

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