Recent advancements in vision-language models (VLMs) have expanded their potential for real-world applications, enabling these models to perform complex reasoning on images. In the widely used fully autoregressive transformer-based models like LLaVA, projected visual tokens are prepended to textual tokens. Oftentimes, visual tokens are significantly more than prompt tokens, resulting in increased computational overhead during both training and inference. In this paper, we propose Visual Compact Token Registers (Victor), a method that reduces the number of visual tokens by summarizing them into a smaller set of register tokens. Victor adds a few learnable register tokens after the visual tokens and summarizes the visual information into these registers using the first few layers in the language tower of VLMs. After these few layers, all visual tokens are discarded, significantly improving computational efficiency for both training and inference. Notably, our method is easy to implement and requires a small number of new trainable parameters with minimal impact on model performance. In our experiment, with merely 8 visual registers--about 1% of the original tokens--Victor shows less than a 4% accuracy drop while reducing the total training time by 43% and boosting the inference throughput by 3.3 . Vision-language models (VLMs) have attracted considerable attention for their capability to process visual and textual information, enabling various real-world applications, such as image captioning, visual question answering, and multimodal reasoning (OpenAI, 2023; Liu et al., 2024c). For example, GPT-4V (OpenAI, 2023) demonstrates the potential of these models in helping visually impaired individuals to "see" the world through cell phone cameras.

The proliferation of creative video content has driven demand for textual descriptions or summaries that allow users to recall key plot points or get an overview without watching. The volume of movie content and speed of turnover motivates automatic summarisation, which is nevertheless challenging, requiring identifying character intentions and very long-range temporal dependencies. The few existing methods attempting this task rely heavily on textual screenplays as input, greatly limiting their applicability. In this work, we propose the task of automatic screenplay generation, and a method, ScreenWriter, that operates only on video and produces output which includes dialogue, speaker names, scene breaks, and visual descriptions. ScreenWriter introduces a novel algorithm to segment the video into scenes based on the sequence of visual vectors, and a novel method for the challenging problem of determining character names, based on a database of actors' faces. We further demonstrate how these automatic screenplays can be used to generate plot synopses with a hierarchical summarisation method based on scene breaks. We test the quality of the final summaries on the recent MovieSum dataset, which we augment with videos, and show that they are superior to a number of comparison models which assume access to goldstandard screenplays.

With the rapid development of the internet, the richness of User-Generated Contentcontinues to increase, making Multimodal Aspect-Based Sentiment Analysis (MABSA) a research hotspot. Existing studies have achieved certain results in MABSA, but they have not effectively addressed the analytical challenges in scenarios where multiple entities and sentiments coexist. This paper innovatively introduces Large Language Models (LLMs) for event decomposition and proposes a reinforcement learning framework for Multimodal Aspect-based Sentiment Analysis (MABSA-RL) framework. This framework decomposes the original text into a set of events using LLMs, reducing the complexity of analysis, introducing reinforcement learning to optimize model parameters. Experimental results show that MABSA-RL outperforms existing advanced methods on two benchmark datasets. This paper provides a new research perspective and method for multimodal aspect-level sentiment analysis.

Causal networks are widely used in many fields to model the complex relationships between variables. A recent approach has sought to construct causal networks by leveraging the wisdom of crowds through the collective participation of humans. While this can yield detailed causal networks that model the underlying phenomena quite well, it requires a large number of individuals with domain understanding. We adopt a different approach: leveraging the causal knowledge that large language models, such as OpenAI's GPT-4, have learned by ingesting massive amounts of literature. Within a dedicated visual analytics interface, called CausalChat, users explore single variables or variable pairs recursively to identify causal relations, latent variables, confounders, and mediators, constructing detailed causal networks through conversation. Each probing interaction is translated into a tailored GPT-4 prompt and the response is conveyed through visual representations which are linked to the generated text for explanations. We demonstrate the functionality of CausalChat across diverse data contexts and conduct user studies involving both domain experts and laypersons.

Mathematical reasoning has been a challenging aspect of large language models (LLMs). However, the introduction of step-by-step Chain-of-Thought (CoT) inference has significantly advanced the mathematical capabilities of LLMs. Despite this progress, current approaches either require massive inference datasets as training datasets or rely on few-shot methods that often sacrifice accuracy. To address this bottleneck in mathematical reasoning, we propose a novel method called Step Guidance Reasoning without involving further model fine-tuning. In this approach, LLMs reflect on small reasoning steps -- similar to how humans deliberate on and focus attention on what to do next. By incorporating this reflective process into the inference stage, LLMs can effectively guide their reasoning from one step to the next. Our method significantly improved the math performance, raising the accuracy on the AMC23 dataset from 30% to 57.5%, a relative improvement of 91.7%, and on the sampled level 5 problem of the MATH dataset, we achieved a relative accuracy improvement of 55.8%, increasing from 43% to 67%.

Geometry problem-solving demands advanced reasoning abilities to process multimodal inputs and employ mathematical knowledge effectively. Vision-language models (VLMs) have made significant progress in various multimodal tasks. Yet, they still struggle with geometry problems and are significantly limited by their inability to perform mathematical operations not seen during pre-training, such as calculating the cosine of an arbitrary angle, and by difficulties in correctly applying relevant geometry formulas. To overcome these challenges, we present GeoCoder, which leverages modular code-finetuning to generate and execute code using a predefined geometry function library. By executing the code, we achieve accurate and deterministic calculations, contrasting the stochastic nature of autoregressive token prediction, while the function library minimizes errors in formula usage. We also propose a multimodal retrieval-augmented variant of GeoCoder, named RAG-GeoCoder, which incorporates a non-parametric memory module for retrieving functions from the geometry library, thereby reducing reliance on parametric memory. Our modular code-finetuning approach enhances the geometric reasoning capabilities of VLMs, yielding an average improvement of over 16% across various question complexities on the GeomVerse dataset compared to other finetuning methods.

Large language models (LLMs) have demonstrated impressive performance on several tasks and are increasingly deployed in real-world applications. However, especially in high-stakes settings, it becomes vital to know when the output of an LLM may be unreliable. Depending on whether an answer is trustworthy, a system can then choose to route the question to another expert, or otherwise fall back on a safe default behavior. In this work, we study the extent to which LLMs can reliably indicate confidence in their answers, and how this notion of confidence can translate into downstream accuracy gains. We propose Self-REF, a lightweight training strategy to teach LLMs to express confidence in whether their answers are correct in a reliable manner. Self-REF introduces confidence tokens into the LLM, from which a confidence score can be extracted. Compared to conventional approaches such as verbalizing confidence and examining token probabilities, we demonstrate empirically that confidence tokens show significant improvements in downstream routing and rejection learning tasks.

Challenges in managing linguistic diversity and integrating various musical modalities are faced by current music information retrieval systems. These limitations reduce their effectiveness in a global, multimodal music environment. To address these issues, we introduce CLaMP 2, a system compatible with 101 languages that supports both ABC notation (a text-based musical notation format) and MIDI (Musical Instrument Digital Interface) for music information retrieval. CLaMP 2, pre-trained on 1.5 million ABC-MIDI-text triplets, includes a multilingual text encoder and a multimodal music encoder aligned via contrastive learning. By leveraging large language models, we obtain refined and consistent multilingual descriptions at scale, significantly reducing textual noise and balancing language distribution. Our experiments show that CLaMP 2 achieves state-of-the-art results in both multilingual semantic search and music classification across modalities, thus establishing a new standard for inclusive and global music information retrieval.

Assessing the capabilities and limitations of large language models (LLMs) has garnered significant interest, yet the evaluation of multiple models in real-world scenarios remains rare. Multilingual evaluation often relies on translated benchmarks, which typically do not capture linguistic and cultural nuances present in the source language. This study provides an extensive assessment of 24 LLMs on real world data collected from Indian patients interacting with a medical chatbot in Indian English and 4 other Indic languages. We employ a uniform Retrieval Augmented Generation framework to generate responses, which are evaluated using both automated techniques and human evaluators on four specific metrics relevant to our application. We find that models vary significantly in their performance and that instruction tuned Indic models do not always perform well on Indic language queries. Further, we empirically show that factual correctness is generally lower for responses to Indic queries compared to English queries. Finally, our qualitative work shows that code-mixed and culturally relevant queries in our dataset pose challenges to evaluated models.

Summarization is one of the most common tasks performed by large language models (LLMs), especially in applications like Retrieval-Augmented Generation (RAG). However, existing evaluations of hallucinations in LLM-generated summaries, and evaluations of hallucination detection models both suffer from a lack of diversity and recency in the LLM and LLM families considered. This paper introduces FaithBench, a summarization hallucination benchmark comprising challenging hallucinations made by 10 modern LLMs from 8 different families, with ground truth annotations by human experts. ``Challenging'' here means summaries on which popular, state-of-the-art hallucination detection models, including GPT-4o-as-a-judge, disagreed on. Our results show GPT-4o and GPT-3.5-Turbo produce the least hallucinations. However, even the best hallucination detection models have near 50\% accuracies on FaithBench, indicating lots of room for future improvement. The repo is https://github.com/vectara/FaithBench