Research map / Vision-language and multimodal / Multimodal LLMs
Explicit Logic Channel for Validation and Enhancement of MLLMs on Zero-Shot Tasks
Mei Chee Leong, Gu Ying, Hui Tan, Liyuan Li, Nancy Chen
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add aligned fusion of ELC with the MLLM | InternVL2.5-8B · NegBench COCO | accuracy: 0.912 -> 0.965 | Improved |
| add aligned fusion of ELC with the MLLM | InternVL2.0-8B · NegBench COCO | accuracy: 0.4878 -> 0.8434 | Improved |
| add aligned fusion of ELC with the MLLM | LLaVA-1.5-13B · NegBench COCO | accuracy: 0.6206 -> 0.6206 | No clear effect |
| add aligned fusion of ELC with the MLLM | Gemma-3-12B · HC-RefCOCOg test | Acc@0.5: .088 -> .752 | Improved |
| add aligned fusion of ELC with the MLLM | Qwen3.0-VL-8B · HC-RefCOCOg test | Acc@0.5: .818 -> .856 | Improved |
| add aligned fusion of ELC with the MLLM | Gemma-3-12B · HC-RefLoCo test | Acc@0.5: .0920 -> .4270 | Improved |
| add aligned fusion of ELC with the MLLM | Qwen3.0-VL-8B · HC-RefLoCo test | Acc@0.5: .9418 -> .9388 | No clear effect |
| use Qwen3.0 instead of Mistral as ELC language parser (ablation) | ELC · NegBench COCO | ELC accuracy: 0.7577 -> 0.7614 | No clear effect |
| use Qwen3.0 instead of Mistral as ELC language parser (ablation) | ELC · HC-RefCOCOg val | ELC Acc@0.5: 0.8074 -> 0.7996 | No clear effect |
| use Mistral instead of Qwen3.0 as ELC language parser (ablation) | ELC · HC-RefLoCo val | ELC Acc@0.5: .8044 -> .8110 | No clear effect |
| mask one critical object in each image (adversarial) (ablation) | Qwen3.0-VL ILC · HC-RefCOCOg adversarial (100 images) | Acc@0.5: 91% -> 60% | Hurt |
| mask one critical object in each image (adversarial) (ablation) | ELC · HC-RefCOCOg adversarial (100 images) | Acc@0.5: 82% -> 41% | Hurt |
Closest papers by meaning
- Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models (CVPR 2025 · similarity 0.90)
- VisuLogic: A Benchmark for Evaluating Visual Reasoning in Multi-modal Large Language Models (ICLR 2026 · similarity 0.89)
- LENS: Multi-level Evaluation of Multimodal Reasoning with Large Language Models (ICLR 2026 · similarity 0.89)
- VisuRiddles: Fine-grained Perception is a Primary Bottleneck for Multimodal Large Language Models in Abstract Visual Reasoning (ICLR 2026 · similarity 0.89)
- Look-Back: Implicit Visual Re-focusing in MLLM Reasoning (AAAI 2026 · similarity 0.89)
- Vision-aligned Latent Reasoning for Multi-modal Large Language Model (ICML 2026 · similarity 0.89)
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.
