AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Contrastive-Guided Self-Supervised Latent Visual Reasoning for Hallucination Mitigation

Aoxue Dai, Ningning Wang

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Reinforcement learning for visual reasoning · 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
add RL stage with contrastive-guided GRPO after SFT (ablation)CoLVR-3B · POPEF1: 84.4 -> 85.9Improved
add RL stage with contrastive-guided GRPO after SFT (ablation)CoLVR-3B · Hallbenchaccuracy: 40.8 -> 42.6Improved
add RL stage with contrastive-guided GRPO after SFT (ablation)CoLVR-3B · MMHalscore: 4.46 -> 4.78Improved
add RL stage with contrastive-guided GRPO after SFT (ablation)CoLVR-3B · CHAIRCHAIRs: 38.6 -> 30.8Improved
add SFT-only latent training to base model (ablation)Qwen2.5-VL-3B · CHAIRF1: 79.0 -> 78.3No clear effect
add latent-only backpropagation (stop-gradient) (ablation)CoLVR-3B · POPEaccuracy: 84.6 -> 88.2Improved
add latent-only backpropagation (stop-gradient) (ablation)CoLVR-3B · MMHalscore: 4.28 -> 4.78Improved
use Gaussian-noise contrast instead of blank image (ablation)CoLVR-3B · CHAIRCHAIRs: 36.9 -> 30.8Improved
use contrastive saliency instead of original-image attention (ablation)CoLVR-3B · CHAIRCHAIRs: 33.3 -> 30.8Improved
use correct contrast direction instead of reverse contrast (ablation)CoLVR-3B · CHAIRCHAIRs: 46.9 -> 30.8Improved
vary latent length K from 1 to 32 (ablation)CoLVR-3B · CHAIRCHAIRs and F1Mixed
use CoLVR self-supervised instead of Monet supervised latent reasoningCoLVR-7B vs Monet-7B · CHAIRCHAIRs: 34.1 -> 20.2Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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