Raja Giryes


2025

pdf bib
Mamba Knockout for Unraveling Factual Information Flow
Nir Endy | Idan Daniel Grosbard | Yuval Ran-Milo | Yonatan Slutzky | Itay Tshuva | Raja Giryes
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

This paper investigates the flow of factual information in Mamba State-Space Model (SSM)-based language models. We rely on theoretical and empirical connections to Transformer-based architectures and their attention mechanisms. Exploiting this relationship, we adapt attentional interpretability techniques originally developed for Transformers—specifically, the Attention Knockout methodology—to both Mamba-1 and Mamba-2. Using them we trace how information is transmitted and localized across tokens and layers, revealing patterns of subject-token information emergence and layer-wise dynamics. Notably, some phenomena vary between mamba models and Transformer based models, while others appear universally across all models inspected—hinting that these may be inherent to LLMs in general. By further leveraging Mamba’s structured factorization, we disentangle how distinct “features” either enable token-to-token information exchange or enrich individual tokens, thus offering a unified lens to understand Mamba internal operations.

pdf bib
Performance Gap in Entity Knowledge Extraction Across Modalities in Vision Language Models
Ido Cohen | Daniela Gottesman | Mor Geva | Raja Giryes
Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

Vision-language models (VLMs) excel at extracting and reasoning about information from images. Yet, their capacity to leverage internal knowledge about specific entities remains underexplored. This work investigates the disparity in model performance when answering factual questions about an entity described in text versus depicted in an image. Our results reveal a significant accuracy drop — reaching 18% for some models — when the entity is presented visually instead of textually. To study this gap we present PopVQA, a dataset which allows separating entity recognition and question answering, and use it to benchmark several models. We hypothesize that this decline arises from limitations in how information flows from image tokens to query tokens. Thus, we use mechanistic interpretability tools to reveal that, although image tokens are preprocessed by the vision encoder, meaningful information flow from these tokens occurs only in the much deeper layers. Furthermore, critical image processing happens in the language model’s middle layers, allowing few layers for consecutive reasoning, highlighting a potential inefficiency in how the model utilizes its layers for reasoning. These insights shed light on the internal mechanics of VLMs and offer pathways for enhancing their reasoning capabilities. PopVQA can be found at https://huggingface.co/datasets/idoco/PopVQA.

pdf bib
Bridging the Visual Gap: Fine-Tuning Multimodal Models with Knowledge-Adapted Captions
Moran Yanuka | Assaf Ben-Kish | Yonatan Bitton | Idan Szpektor | Raja Giryes
Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)

Recent research increasingly focuses on training vision-language models (VLMs) with long, detailed image captions. However, small-scale VLMs often struggle to balance the richness of these captions with the risk of hallucinating content during fine-tuning. In this paper, we explore how well VLMs adapt to such captions. To quantify caption quality, we propose Decomposed NLI (DNLI), an evaluation framework that breaks down generated captions into individual propositions, assessing each in isolation. This fine-grained analysis reveals a critical balance between capturing descriptive details and preventing hallucinations. Our findings show that simply reducing caption complexity or employing standard data curation techniques does not effectively resolve this issue. To tackle this challenge, we introduce Knowledge Adapted (KnowAda) fine-tuning, a data-centric approach that automatically adapts training data with the model’s existing knowledge and visual understanding. KnowAda minimizes hallucinations while preserving high descriptiveness. We validate this approach across several small-scale VLMs (up to 7B parameters) and dense caption datasets, demonstrating that KnowAda effectively balances hallucination reduction and descriptiveness. Our results show that KnowAda outperforms various baselines in both automatic metrics and human evaluations.

2024

pdf bib
Mitigating Open-Vocabulary Caption Hallucinations
Assaf Ben-Kish | Moran Yanuka | Morris Alper | Raja Giryes | Hadar Averbuch-Elor
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing

While recent years have seen rapid progress in image-conditioned text generation, image captioning still suffers from the fundamental issue of hallucinations, namely, the generation of spurious details that cannot be inferred from the given image. Existing methods largely use closed-vocabulary object lists to mitigate or evaluate hallucinations in image captioning, ignoring the long-tailed nature of hallucinations that occur in practice. To this end, we propose a framework for addressing hallucinations in image captioning in the open-vocabulary setting. Our framework includes a new benchmark, OpenCHAIR, that leverages generative foundation models to evaluate open-vocabulary object hallucinations for image captioning, surpassing the popular and similarly-sized CHAIR benchmark in both diversity and accuracy. Furthermore, to mitigate open-vocabulary hallucinations without using a closed object list, we propose MOCHa, an approach harnessing advancements in reinforcement learning. Our multi-objective reward function explicitly targets the trade-off between fidelity and adequacy in generations without requiring any strong supervision. MOCHa improves a large variety of image captioning models, as captured by our OpenCHAIR benchmark and other existing metrics. We will release our code and models.

pdf bib
ICC : Quantifying Image Caption Concreteness for Multimodal Dataset Curation
Moran Yanuka | Morris Alper | Hadar Averbuch-Elor | Raja Giryes
Findings of the Association for Computational Linguistics: ACL 2024

Web-scale training on paired text-image data is becoming increasingly central to multimodal learning, but is challenged by the highly noisy nature of datasets in the wild. Standard data filtering approaches succeed in removing mismatched text-image pairs, but permit semantically related but highly abstract or subjective text. These approaches lack the fine-grained ability to isolate the most concrete samples that provide the strongest signal for learning in a noisy dataset. In this work, we propose a new metric, Image Caption Concreteness (ICC), that evaluates caption text without an image reference to measure its concreteness and relevancy for use in multimodal learning. Our unsupervised approach leverages strong foundation models for measuring visual-semantic information loss in multimodal representations. We demonstrate that this strongly correlates with human evaluation of concreteness in both single-word and caption-level texts. Moreover, we show that curation using ICC complements existing approaches: It succeeds in selecting the highest quality samples from multimodal web-scale datasets to allow for efficient training in resource-constrained settings.

2020

pdf bib
Low Resource Sequence Tagging using Sentence Reconstruction
Tal Perl | Sriram Chaudhury | Raja Giryes
Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics

This work revisits the task of training sequence tagging models with limited resources using transfer learning. We investigate several proposed approaches introduced in recent works and suggest a new loss that relies on sentence reconstruction from normalized embeddings. Specifically, our method demonstrates how by adding a decoding layer for sentence reconstruction, we can improve the performance of various baselines. We show improved results on the CoNLL02 NER and UD 1.2 POS datasets and demonstrate the power of the method for transfer learning with low-resources achieving 0.6 F1 score in Dutch using only one sample from it.