Enhancing Multimodal LLM for Detailed and Accurate Video Captioning using Multi-Round Preference Optimization

Videos contain a wealth of information, and generating detailed and accurate descriptions in natural language is a key aspect of video understanding. In this paper, we present video-SALMONN 2, an advanced audio-visual large language model (LLM) with low-rank adaptation (LoRA) designed for enhanced video (with paired audio) captioning through directed preference optimization (DPO). We propose new metrics to evaluate the completeness and accuracy of video descriptions, which are optimized using DPO. To further improve training, we introduce a novel multi-round DPO (mrDPO) approach, which involves periodically updating the DPO reference model, merging and re-initializing the LoRA module as a proxy for parameter updates after each training round (1,000 steps), and incorporating guidance from ground-truth video captions to stabilize the process. To address potential catastrophic forgetting of non-captioning abilities due to mrDPO, we propose rebirth tuning, which finetunes the pre-DPO LLM by using the captions generated by the mrDPO-trained model as supervised labels. Experiments show that mrDPO significantly enhances video-SALMONN 2's captioning accuracy, reducing global and local error rates by 40\% and 20\%, respectively, while decreasing the repetition rate by 35\%. The final video-SALMONN 2 model, with just 7 billion parameters, surpasses leading models such as GPT-4o and Gemini-1.5-Pro in video captioning tasks, while maintaining competitive performance to the state-of-the-art on widely used video question-answering benchmark among models of similar size. Upon acceptance, we will release the code, model checkpoints, and training and test data. Demos are available at \href{https://video-salmonn-2.github.io}{https://video-salmonn-2.github.io}.

Editing Music with Melody and Text: Using ControlNet for Diffusion Transformer

Despite the significant progress in controllable music generation and editing, challenges remain in the quality and length of generated music due to the use of Mel-spectrogram representations and UNet-based model structures. To address these limitations, we propose a novel approach using a Diffusion Transformer (DiT) augmented with an additional control branch using ControlNet. This allows for long-form and variable-length music generation and editing controlled by text and melody prompts. For more precise and fine-grained melody control, we introduce a novel top-$k$ constant-Q Transform representation as the melody prompt, reducing ambiguity compared to previous representations (e.g., chroma), particularly for music with multiple tracks or a wide range of pitch values. To effectively balance the control signals from text and melody prompts, we adopt a curriculum learning strategy that progressively masks the melody prompt, resulting in a more stable training process. Experiments have been performed on text-to-music generation and music-style transfer tasks using open-source instrumental recording data. The results demonstrate that by extending StableAudio, a pre-trained text-controlled DiT model, our approach enables superior melody-controlled editing while retaining good text-to-music generation performance. These results outperform a strong MusicGen baseline in terms of both text-based generation and melody preservation for editing. Audio examples can be found at https://stable-audio-control.github.io/web/.

LoRC: Low-Rank Compression for LLMs KV Cache with a Progressive Compression Strategy

The Key-Value (KV) cache is a crucial component in serving transformer-based autoregressive large language models (LLMs), enabling faster inference by storing previously computed KV vectors. However, its memory consumption scales linearly with sequence length and batch size, posing a significant bottleneck in LLM deployment. Existing approaches to mitigate this issue include: (1) efficient attention variants integrated in upcycling stages, which requires extensive parameter tuning thus unsuitable for pre-trained LLMs; (2) KV cache compression at test time, primarily through token eviction policies, which often overlook inter-layer dependencies and can be task-specific. This paper introduces an orthogonal approach to KV cache compression. We propose a low-rank approximation of KV weight matrices, allowing for plug-in integration with existing transformer-based LLMs without model retraining. To effectively compress KV cache at the weight level, we adjust for layerwise sensitivity and introduce a progressive compression strategy, which is supported by our theoretical analysis on how compression errors accumulate in deep networks. Our method is designed to function without model tuning in upcycling stages or task-specific profiling in test stages. Extensive experiments with LLaMA models ranging from 8B to 70B parameters across various tasks show that our approach significantly reduces the GPU memory footprint while maintaining performance.

SWIM: Short-Window CNN Integrated with Mamba for EEG-Based Auditory Spatial Attention Decoding

In complex auditory environments, the human auditory system possesses the remarkable ability to focus on a specific speaker while disregarding others. In this study, a new model named SWIM, a short-window convolution neural network (CNN) integrated with Mamba, is proposed for identifying the locus of auditory attention (left or right) from electroencephalography (EEG) signals without relying on speech envelopes. SWIM consists of two parts. The first is a short-window CNN (SW$_\text{CNN}$), which acts as a short-term EEG feature extractor and achieves a final accuracy of 84.9% in the leave-one-speaker-out setup on the widely used KUL dataset. This improvement is due to the use of an improved CNN structure, data augmentation, multitask training, and model combination. The second part, Mamba, is a sequence model first applied to auditory spatial attention decoding to leverage the long-term dependency from previous SW$_\text{CNN}$ time steps. By joint training SW$_\text{CNN}$ and Mamba, the proposed SWIM structure uses both short-term and long-term information and achieves an accuracy of 86.2%, which reduces the classification errors by a relative 31.0% compared to the previous state-of-the-art result. The source code is available at https://github.com/windowso/SWIM-ASAD.

LW2G: Learning Whether to Grow for Prompt-based Continual Learning

Continual Learning (CL) aims to learn in non-stationary scenarios, progressively acquiring and maintaining knowledge from sequential tasks. Recent Prompt-based Continual Learning (PCL) has achieved remarkable performance with Pre-Trained Models (PTMs). These approaches grow a prompt sets pool by adding a new set of prompts when learning each new task (\emph{prompt learning}) and adopt a matching mechanism to select the correct set for each testing sample (\emph{prompt retrieval}). Previous studies focus on the latter stage by improving the matching mechanism to enhance Prompt Retrieval Accuracy (PRA). To promote cross-task knowledge facilitation and form an effective and efficient prompt sets pool, we propose a plug-in module in the former stage to \textbf{Learn Whether to Grow (LW2G)} based on the disparities between tasks. Specifically, a shared set of prompts is utilized when several tasks share certain commonalities, and a new set is added when there are significant differences between the new task and previous tasks. Inspired by Gradient Projection Continual Learning, our LW2G develops a metric called Hinder Forward Capability (HFC) to measure the hindrance imposed on learning new tasks by surgically modifying the original gradient onto the orthogonal complement of the old feature space. With HFC, an automated scheme Dynamic Growing Approach adaptively learns whether to grow with a dynamic threshold. Furthermore, we design a gradient-based constraint to ensure the consistency between the updating prompts and pre-trained knowledge, and a prompts weights reusing strategy to enhance forward transfer. Extensive experiments show the effectiveness of our method. The source codes are available at \url{https://github.com/RAIAN08/LW2G}.

Enabling Auditory Large Language Models for Automatic Speech Quality Evaluation

Speech quality assessment typically requires evaluating audio from multiple aspects, such as mean opinion score (MOS) and speaker similarity (SIM) etc., which can be challenging to cover using one small model designed for a single task. In this paper, we propose leveraging recently introduced auditory large language models (LLMs) for automatic speech quality assessment. By employing task-specific prompts, auditory LLMs are finetuned to predict MOS, SIM and A/B testing results, which are commonly used for evaluating text-to-speech systems. Additionally, the finetuned auditory LLM is able to generate natural language descriptions assessing aspects like noisiness, distortion, discontinuity, and overall quality, providing more interpretable outputs. Extensive experiments have been performed on the NISQA, BVCC, SOMOS and VoxSim speech quality datasets, using open-source auditory LLMs such as SALMONN, Qwen-Audio, and Qwen2-Audio. For the natural language descriptions task, a commercial model Google Gemini 1.5 Pro is also evaluated. The results demonstrate that auditory LLMs achieve competitive performance compared to state-of-the-art task-specific small models in predicting MOS and SIM, while also delivering promising results in A/B testing and natural language descriptions. Our data processing scripts and finetuned model checkpoints will be released upon acceptance.

MT2KD: Towards A General-Purpose Encoder for Speech, Speaker, and Audio Events

With the advances in deep learning, the performance of end-to-end (E2E) single-task models for speech and audio processing has been constantly improving. However, it is still challenging to build a general-purpose model with high performance on multiple tasks, since different speech and audio processing tasks usually require different training data, input features, or model architectures to achieve optimal performance. In this work, MT2KD, a novel two-stage multi-task learning framework is proposed to build a general-purpose speech and audio encoder that jointly performs three fundamental tasks: automatic speech recognition (ASR), audio tagging (AT) and speaker verification (SV). In the first stage, multi-teacher knowledge distillation (KD) is applied to align the feature spaces of three single-task high-performance teacher encoders into a single student encoder using the same unlabelled data. In the second stage, multi-task supervised fine-tuning is carried out by initialising the model from the first stage and training on the separate labelled data of each single task. Experiments demonstrate that the proposed multi-task training pipeline significantly outperforms a baseline model trained with multi-task learning from scratch. The final system achieves good performance on ASR, AT and SV: with less than 4% relative word-error-rate increase on ASR, only 1.9 lower mean averaged precision on AT and 0.23% absolute higher equal error rate on SV compared to the best-performing single-task encoders, using only a 66M total model parameters.

Beyond Relevance: Improving User Engagement by Personalization for Short-Video Search

Personalized search has been extensively studied in various applications, including web search, e-commerce, social networks, etc. With the soaring popularity of short-video platforms, exemplified by TikTok and Kuaishou, the question arises: can personalization elevate the realm of short-video search, and if so, which techniques hold the key? In this work, we introduce $\text{PR}^2$, a novel and comprehensive solution for personalizing short-video search, where $\text{PR}^2$ stands for the Personalized Retrieval and Ranking augmented search system. Specifically, $\text{PR}^2$ leverages query-relevant collaborative filtering and personalized dense retrieval to extract relevant and individually tailored content from a large-scale video corpus. Furthermore, it utilizes the QIN (Query-Dominate User Interest Network) ranking model, to effectively harness user long-term preferences and real-time behaviors, and efficiently learn from user various implicit feedback through a multi-task learning framework. By deploying the $\text{PR}^2$ in production system, we have achieved the most remarkable user engagement improvements in recent years: a 10.2% increase in CTR@10, a notable 20% surge in video watch time, and a 1.6% uplift of search DAU. We believe the practical insights presented in this work are valuable especially for building and improving personalized search systems for the short video platforms.

Large Language Model Based Generative Error Correction: A Challenge and Baselines for Speech Recognition, Speaker Tagging, and Emotion Recognition

Given recent advances in generative AI technology, a key question is how large language models (LLMs) can enhance acoustic modeling tasks using text decoding results from a frozen, pretrained automatic speech recognition (ASR) model. To explore new capabilities in language modeling for speech processing, we introduce the generative speech transcription error correction (GenSEC) challenge. This challenge comprises three post-ASR language modeling tasks: (i) post-ASR transcription correction, (ii) speaker tagging, and (iii) emotion recognition. These tasks aim to emulate future LLM-based agents handling voice-based interfaces while remaining accessible to a broad audience by utilizing open pretrained language models or agent-based APIs. We also discuss insights from baseline evaluations, as well as lessons learned for designing future evaluations.

Extract and Diffuse: Latent Integration for Improved Diffusion-based Speech and Vocal Enhancement

Diffusion-based generative models have recently achieved remarkable results in speech and vocal enhancement due to their ability to model complex speech data distributions. While these models generalize well to unseen acoustic environments, they may not achieve the same level of fidelity as the discriminative models specifically trained to enhance particular acoustic conditions. In this paper, we propose Ex-Diff, a novel score-based diffusion model that integrates the latent representations produced by a discriminative model to improve speech and vocal enhancement, which combines the strengths of both generative and discriminative models. Experimental results on the widely used MUSDB dataset show relative improvements of 3.7% in SI-SDR and 10.0% in SI-SIR compared to the baseline diffusion model for speech and vocal enhancement tasks, respectively. Additionally, case studies are provided to further illustrate and analyze the complementary nature of generative and discriminative models in this context.