AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Cat-PO: Cross-modal Adaptive Token-rewards for Preference Optimization in Truthful Multimodal LLMs

Zhixiao Zheng, Zheren Fu, Zhiyuan Yao, Dongming Zhang, Zhendong Mao

ICLR 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Unified any-modality language models

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes Cat-PO, a cross-modal adaptive token-reward preference optimization method to reduce hallucinations in multimodal LLMs.

Problem

MLLMs hallucinate, producing text semantically inconsistent with input images; existing multimodal preference optimization methods show shortcomings at the preference data decoding stage, where different response tokens contribute differently.

Approach

Assigns adaptive token-level rewards that account for cross-modal (image-text) relevance of response tokens within preference optimization, rather than treating all tokens uniformly.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add cross-modal token rewards and KL loss to DPO (Cat-PO)LLaVA-v1.5-7B · MM-Halhallucination rate (%): 58.3 -> 42.0 (DPO to Cat-PO)Improved
add cross-modal token rewards and KL loss to DPOLLaVA-v1.5-7B · AMBER-GenerationCHAIR: 5.7 -> 4.8Improved
replace DPO with Cat-POLLaVA-v1.5-7B · AMBER-DiscriminationAccuracy: 77.5 -> 78.0No clear effect
replace DPO with Cat-POLLaVA-v1.5-13B · MM-HalScore: 2.47 -> 2.85Improved
use only attention-based token weights (ablation)Cat-PO · MM-Halhallucination rate (%): 58.3 -> 55.0 (vs DPO-only)Improved
use only semantic-similarity token weights (ablation)Cat-PO · MM-Halhallucination rate (%): 58.3 -> 47.0 (vs DPO-only)Improved
combine attention and similarity weights vs similarity only (ablation)Cat-PO · MM-Halhallucination rate (%): 47.0 -> 42.0 (with KL)Improved
add token-level KL term (ablation)Cat-PO · MM-Halhallucination rate (%): 53.0 -> 42.0Improved
weight top 30% tokens instead of bottom 30% (ablation)Cat-PO · MM-HalMM-Hal score and rateImproved
make fusion weights learnable (ablation)Cat-PO · MM-HalScore: 2.74 -> 2.55; rate 42% -> 50%Hurt
evaluate on POPE adversarial subset versus averageCat-PO · POPEF1: 84.0 -> 82.5 (-2%)No clear effect

Experimental setup

Builds on: direct preference optimization

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