DeP mitigates MLLM hallucinations by dynamically perturbing text prompts to identify and reinforce stable visual evidence regions while counteracting language prior biases using attention variance and logit statistics.
Title resolution pending
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
baseline 1
dataset 1
citation-polarity summary
representative citing papers
DORI shows state-of-the-art multimodal models perform near-random on fine-grained object orientation reasoning despite strong scores on broader spatial benchmarks.
MM1 models achieve state-of-the-art few-shot multimodal results by pre-training on a careful mix of image-caption, interleaved, and text-only data with optimized image encoders.
ESC uses emotional cues triggered by an external verifier to enable training-free self-correction in VLMs, improving reliability on safety, hallucination, and reasoning benchmarks.
Falcon introduces a structured intermediate safety state for compositional threat reasoning in X-ray baggage screening and a benchmark Falcon-X to evaluate it.
A vision-language model pre-trained via instruction tuning on CT-report pairs improves survival prediction accuracy over baselines, especially when clinical data alone is weak, while also producing text answers to clinical questions.
A multilingual VLM training corpus and evaluation benchmark suite for five European languages shows consistent cross-lingual transfer gains without English degradation.
citing papers explorer
-
Decoding by Perturbation: Mitigating MLLM Hallucinations via Dynamic Textual Perturbation
DeP mitigates MLLM hallucinations by dynamically perturbing text prompts to identify and reinforce stable visual evidence regions while counteracting language prior biases using attention variance and logit statistics.
-
Seeing Isn't Orienting: A Cognitively Informed Hierarchical Benchmark for Object Orientation in MLLMs
DORI shows state-of-the-art multimodal models perform near-random on fine-grained object orientation reasoning despite strong scores on broader spatial benchmarks.
-
MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training
MM1 models achieve state-of-the-art few-shot multimodal results by pre-training on a careful mix of image-caption, interleaved, and text-only data with optimized image encoders.
-
ESC: Emotional Self-Correction for Reliable Vision-Language Models
ESC uses emotional cues triggered by an external verifier to enable training-free self-correction in VLMs, improving reliability on safety, hallucination, and reasoning benchmarks.
-
Falcon: Functional Assembly and Language for Compositional Reasoning in X-ray
Falcon introduces a structured intermediate safety state for compositional threat reasoning in X-ray baggage screening and a benchmark Falcon-X to evaluate it.
-
Medical Image Understanding Improves Survival Prediction via Visual Instruction Tuning
A vision-language model pre-trained via instruction tuning on CT-report pairs improves survival prediction accuracy over baselines, especially when clinical data alone is weak, while also producing text answers to clinical questions.
-
Multilingual Training and Evaluation Resources for Vision-Language Models
A multilingual VLM training corpus and evaluation benchmark suite for five European languages shows consistent cross-lingual transfer gains without English degradation.