Daniel Tang

Zhifei Xie, Daniel Tang, Dingwei Tan et al.

Current video generation models excel at creating short, realistic clips, but struggle with longer, multi-scene videos. We introduce \texttt{DreamFactory}, an LLM-based framework that tackles this challenge. \texttt{DreamFactory} leverages multi-agent collaboration principles and a Key Frames Iteration Design Method to ensure consistency and style across long videos. It utilizes Chain of Thought (COT) to address uncertainties inherent in large language models. \texttt{DreamFactory} generates long, stylistically coherent, and complex videos. Evaluating these long-form videos presents a challenge. We propose novel metrics such as Cross-Scene Face Distance Score and Cross-Scene Style Consistency Score. To further research in this area, we contribute the Multi-Scene Videos Dataset containing over 150 human-rated videos.

Shi Wang, Daniel Tang, Luchen Zhang et al.

International Classification of Diseases (ICD) is a set of classification codes for medical records. Automated ICD coding, which assigns unique International Classification of Diseases codes with each medical record, is widely used recently for its efficiency and error-prone avoidance. However, there are challenges that remain such as heterogeneity, label unbalance, and complex relationships between ICD codes. In this work, we proposed a novel Bidirectional Hierarchy Framework(HieNet) to address the challenges. Specifically, a personalized PageRank routine is developed to capture the co-relation of codes, a bidirectional hierarchy passage encoder to capture the codes' hierarchical representations, and a progressive predicting method is then proposed to narrow down the semantic searching space of prediction. We validate our method on two widely used datasets. Experimental results on two authoritative public datasets demonstrate that our proposed method boosts state-of-the-art performance by a large margin.

Xiaochen Zhang, Daniel Tang

We leverage different context windows when predicting the emotion of different utterances. New modules are included to realize variable-length context: 1) two speaker-aware units, which explicitly model inner- and inter-speaker dependencies to form distilled conversational context, and 2) a top-k normalization layer, which determines the most proper context windows from the conversational context to predict emotion. Experiments and ablation studies show that our approach outperforms several strong baselines on three public datasets.

Xinshi Wang, Daniel Tang

Medical dialogue information extraction is becoming an increasingly significant problem in modern medical care. It is difficult to extract key information from electronic medical records (EMRs) due to their large numbers. Previously, researchers proposed attention-based models for retrieving features from EMRs, but their limitations were reflected in their inability to recognize different categories in medical dialogues. In this paper, we propose a novel model, Expert System and Attention for Labelling (ESAL). We use mixture of experts and pre-trained BERT to retrieve the semantics of different categories, enabling the model to fuse the differences between them. In our experiment, ESAL was applied to a public dataset and the experimental results indicated that ESAL significantly improved the performance of Medical Information Classification.

Yewei Song, Cedric Lothritz, Daniel Tang et al.

This paper revisits recent code similarity evaluation metrics, particularly focusing on the application of Abstract Syntax Tree (AST) editing distance in diverse programming languages. In particular, we explore the usefulness of these metrics and compare them to traditional sequence similarity metrics. Our experiments showcase the effectiveness of AST editing distance in capturing intricate code structures, revealing a high correlation with established metrics. Furthermore, we explore the strengths and weaknesses of AST editing distance and prompt-based GPT similarity scores in comparison to BLEU score, execution match, and Jaccard Similarity. We propose, optimize, and publish an adaptable metric that demonstrates effectiveness across all tested languages, representing an enhanced version of Tree Similarity of Edit Distance (TSED).

Ziqi Zhong, Daniel Tang

While multi-modal learning has advanced significantly, current approaches often treat modalities separately, creating inconsistencies in representation and reasoning. We introduce MANTA (Multi-modal Abstraction and Normalization via Textual Alignment), a theoretically-grounded framework that unifies visual and auditory inputs into a structured textual space for seamless processing with large language models. MANTA addresses four key challenges: (1) semantic alignment across modalities with information-theoretic optimization, (2) adaptive temporal synchronization for varying information densities, (3) hierarchical content representation for multi-scale understanding, and (4) context-aware retrieval of sparse information from long sequences. We formalize our approach within a rigorous mathematical framework, proving its optimality for context selection under token constraints. Extensive experiments on the challenging task of Long Video Question Answering show that MANTA improves state-of-the-art models by up to 22.6% in overall accuracy, with particularly significant gains (27.3%) on videos exceeding 30 minutes. Additionally, we demonstrate MANTA's superiority on temporal reasoning tasks (23.8% improvement) and cross-modal understanding (25.1% improvement). Our framework introduces novel density estimation techniques for redundancy minimization while preserving rare signals, establishing new foundations for unifying multimodal representations through structured text.

Daniel Tang

Agent-based models are a powerful tool for studying the behaviour of complex systems that can be described in terms of multiple, interacting ``agents''. However, because of their inherently discrete and often highly non-linear nature, it is very difficult to reason about the relationship between the state of the model, on the one hand, and our observations of the real world on the other. In this paper we consider agents that have a discrete set of states that, at any instant, act with a probability that may depend on the environment or the state of other agents. Given this, we show how the mathematical apparatus of quantum field theory can be used to reason probabilistically about the state and dynamics the model, and describe an algorithm to update our belief in the state of the model in the light of new, real-world observations. Using a simple predator-prey model on a 2-dimensional spatial grid as an example, we demonstrate the assimilation of incomplete, noisy observations and show that this leads to an increase in the mutual information between the actual state of the observed system and the posterior distribution given the observations, when compared to a null model.