arXiv subjects

Fengji Ma

Publications and source records attributed to Fengji Ma.

8 recordsLinked to original sources

arXiv2026

Closing the Verification Loop: Self-Check Captioning for Long-Paragraph Detailed Audio Captioning

Long-paragraph detailed audio captioning, which requires dense and transcript-faithful descriptions of fine-grained audio content, remains unsolved for current audio-visual multimodal language models. We attribute this failure to two structural problems. The first is data poverty, as no public corpus jointly provides long clips, paragraph captions, and verbatim-transcript fidelity. The second is generation-mode failure, evidenced by a 44.8 to 46.4 percentage-point gap between right-audio and shuffled-audio multiple-choice question (MCQ) accuracy. We address both within Self-Check Captioning (SCC), a unified framework that instantiates audio-grounded question answering as the verification primitive at every lifecycle stage. SCC yields three artifacts. Long-paragraph Audio Caption 50k (LACap-50k) is a 50,222-clip audio-visual corpus with 491.5-word captions and a post-hoc automatic speech recognition (ASR) audit. Layer-Curvature Supervised Fine-Tuning (LC-SFT) is the first on-policy supervised fine-tuning method to weight tokens by intermediate-layer evidence, motivated by our identification of Late-Layer Semantic-Entropy Collapse (SEC). SCC-Verifier arbitrates among caption rollouts via audio-grounded self-answering at inference. Across multiple benchmarks, our system attains state-of-the-art among open-source captioners and is competitive with proprietary baselines. We release LACap-50k to fill the resource gap for long-paragraph detailed audio captioning research.

cs.SD↗

arXiv2026

ACE-Cap: Active Evidence Acquisition via Agentic Co-Evolution for Long-Paragraph Fine-Grained Audio Captioning

Long-paragraph fine-grained audio captioning requires models to recover diverse acoustic facts while avoiding omissions and unsupported details. However, prevailing captioners remain passive one-shot generators: once a detail is overlooked, they cannot identify the evidence gap, query the audio for targeted information, or decide when sufficient evidence has been collected. We formulate this task as active evidence acquisition and introduce Agentic Co-Evolution for Captioning (ACE-Cap). The framework uses multi-turn interaction between a Composer and an Instruct model to form a closed evidence-acquisition loop. A Captioner first produces an initial description. Conditioned on this description and the interaction history, a text-only Composer asks targeted questions about unresolved acoustic attributes, while an audio-conditioned Instruct model provides grounded answers. The Composer then decides when to terminate and synthesizes the accumulated evidence into a final caption. ACE-Cap trains these roles through a unified gold-to-prediction reward derived from fixed, gold-grounded multiple-choice questions and a frozen caption-only judge. For credit assignment in variable-length interactions, LOOP-GRPO replaces the trajectory-wide scalar advantage with span-aligned signals: leave-one-out contributions of individual questions to the accumulated evidence, a quality-cost utility for stopping, and an evidence-preservation utility for final synthesis. Role-wise warm-up followed by alternating Composer and Instruct optimization keeps each update a well-defined single-policy problem while allowing the roles to co-evolve. ACE-Cap thus turns captioning from passive one-shot generation into an adaptive process that learns what evidence to acquire, when to stop, and how to preserve it in a long-paragraph caption.

cs.SD↗

arXiv2026

AudioMap: Cloze-and-Choice Reinforcement Learning for Time-Aware Dense Audio Captioning

Time-aware dense audio captioning (TDAC) aims to generate multiple fine-grained attributes (dense) of the audio with precise time boundaries (time-aware). Existing methods struggle to achieve these two goals and mainly rely on supervised fine-tuning, yielding sub-optimal performance. While reinforcement learning (RL) shows promise, applying it to TDAC faces two main challenges: (1) existing rewards are too coarse to supervise multi-event, multi-attribute, and multi-relation descriptions in a fine-grained manner; and (2) temporal supervision is difficult for free-form captions, where flexible event-time expressions make reliable event-time correspondence challenging. To address these challenges, we propose AudioMap, a novel RL-based TDAC framework, which shifts to a unified cloze-and-choice reward paradigm. Specifically, we introduce the Evidence Sufficiency Reward (ESR) with an asymmetric hierarchical scoring mechanism to promote fine-grained accuracy and descriptive richness across diverse acoustic dimensions. Furthermore, we design the Event-Conditioned Temporal Reward (ECTR) to structurally bind timestamps to event semantics via temporal IoU, accompanied by a dual-curriculum learning strategy to facilitate the training process. Finally, to support this task, we construct the first time-aware fine-grained audio captioning dataset, AudioMapCap-44K, which contains 44K carefully annotated captions. Extensive experiments across diverse benchmarks show that AudioMap achieves state-of-the-art (SOTA) performance among open-source models and delivers competitive or superior results relative to proprietary models. Project page and release updates are available at https://github.com/ryysayhi/AudioMap.

cs.SD↗

arXiv2025

PhyAVBench: A Challenging Audio Physics-Sensitivity Benchmark for Physically Grounded Text-to-Audio-Video Generation

Text-to-audio-video (T2AV) generation is central to applications such as filmmaking and world modeling. However, current models often fail to produce physically plausible sounds. Previous benchmarks primarily focus on audio-video temporal synchronization, while largely overlooking explicit evaluation of audio-physics grounding, thereby limiting the study of physically plausible audio-visual generation. To address this issue, we present PhyAVBench, the first benchmark that systematically evaluates the audio-physics grounding capabilities of T2AV, image-to-audio-video (I2AV), and video-to-audio (V2A) models. PhyAVBench offers PhyAV-Sound-11K, a new dataset of 25.5 hours of 11,605 audible videos collected from 184 participants to ensure diversity and avoid data leakage. It contains 337 paired-prompt groups with controlled physical variations that drive sound differences, each grounded with an average of 17 videos and spanning 6 audio-physics dimensions and 41 fine-grained test points. Each prompt pair is annotated with the physical factors underlying their acoustic differences. Importantly, PhyAVBench leverages paired text prompts to evaluate this capability. We term this evaluation paradigm the Audio-Physics Sensitivity Test (APST) and introduce a novel metric, the Contrastive Physical Response Score (CPRS), which quantifies the acoustic consistency between generated videos and their real-world counterparts. We conduct a comprehensive evaluation of 17 state-of-the-art models. Our results reveal that even leading commercial models struggle with fundamental audio-physical phenomena, exposing a critical gap beyond audio-visual synchronization and pointing to future research directions. We hope PhyAVBench will serve as a foundation for advancing physically grounded audio-visual generation. Prompts, ground-truth, and generated video samples are available at https://github.com/imxtx/PhyAVBench.

cs.SD↗

arXiv2025

Quantum-Enhanced Multi-Task Learning with Learnable Weighting for Pharmacokinetic and Toxicity Prediction

Prediction for ADMET (Absorption, Distribution, Metabolism, Excretion, and Toxicity) plays a crucial role in drug discovery and development, accelerating the screening and optimization of new drugs. Existing methods primarily rely on single-task learning (STL), which often fails to fully exploit the complementarities between tasks. Besides, it requires more computational resources while training and inference of each task independently. To address these issues, we propose a new unified Quantum-enhanced and task-Weighted Multi-Task Learning (QW-MTL) framework, specifically designed for ADMET classification tasks. Built upon the Chemprop-RDKit backbone, QW-MTL adopts quantum chemical descriptors to enrich molecular representations with additional information about the electronic structure and interactions. Meanwhile, it introduces a novel exponential task weighting scheme that combines dataset-scale priors with learnable parameters to achieve dynamic loss balancing across tasks. To the best of our knowledge, this is the first work to systematically conduct joint multi-task training across all 13 Therapeutics Data Commons (TDC) classification benchmarks, using leaderboard-style data splits to ensure a standardized and realistic evaluation setting. Extensive experimental results show that QW-MTL significantly outperforms single-task baselines on 12 out of 13 tasks, achieving high predictive performance with minimal model complexity and fast inference, demonstrating the effectiveness and efficiency of multi-task molecular learning enhanced by quantum-informed features and adaptive task weighting.

cs.LG↗

arXiv2024

A Comprehensive Survey on Human Video Generation: Challenges, Methods, and Insights

Human video generation is a dynamic and rapidly evolving task that aims to synthesize 2D human body video sequences with generative models given control conditions such as text, audio, and pose. With the potential for wide-ranging applications in film, gaming, and virtual communication, the ability to generate natural and realistic human video is critical. Recent advancements in generative models have laid a solid foundation for the growing interest in this area. Despite the significant progress, the task of human video generation remains challenging due to the consistency of characters, the complexity of human motion, and difficulties in their relationship with the environment. This survey provides a comprehensive review of the current state of human video generation, marking, to the best of our knowledge, the first extensive literature review in this domain. We start with an introduction to the fundamentals of human video generation and the evolution of generative models that have facilitated the field's growth. We then examine the main methods employed for three key sub-tasks within human video generation: text-driven, audio-driven, and pose-driven motion generation. These areas are explored concerning the conditions that guide the generation process. Furthermore, we offer a collection of the most commonly utilized datasets and the evaluation metrics that are crucial in assessing the quality and realism of generated videos. The survey concludes with a discussion of the current challenges in the field and suggests possible directions for future research. The goal of this survey is to offer the research community a clear and holistic view of the advancements in human video generation, highlighting the milestones achieved and the challenges that lie ahead.

cs.CV↗

arXiv2024

TIMA: Text-Image Mutual Awareness for Balancing Zero-Shot Adversarial Robustness and Generalization Ability

Achieving zero-shot adversarial robustness without sacrificing generalization remains challenging for foundation models such as CLIP, especially under large adversarial perturbations. Through empirical analyses, we identify three critical yet overlooked issues: (1) Logit margins exhibit a stable offset between small and large adversarial perturbations, suggesting that explicitly adjusting margins could improve robustness against unseen large perturbations. (2) A significant negative correlation exists between logit margin and inter-class semantic similarity, indicating that semantic structures are insufficiently leveraged by existing methods. (3) Existing methods for adjusting text embeddings disrupt the intrinsic semantic consistency established by pre-trained models, undermining generalization capability. Motivated by these findings, we propose a novel Text-Image Mutual Awareness (TIMA) framework, including a Text-Aware Image (TAI) tuning module with an Adaptive Semantic-Aware Margin (ASAM) to explicitly calibrate logit margins, and an Image-Aware Text (IAT) tuning module with Semantic Consistent Minimum Hyperspherical Energy (SC-MHE) to preserve semantic consistency. Comprehensive experiments validate that TIMA significantly outperforms existing approaches by effectively addressing the identified limitations.

cs.CV↗

arXiv2023

A Survey on Deep Multi-modal Learning for Body Language Recognition and Generation

Body language (BL) refers to the non-verbal communication expressed through physical movements, gestures, facial expressions, and postures. It is a form of communication that conveys information, emotions, attitudes, and intentions without the use of spoken or written words. It plays a crucial role in interpersonal interactions and can complement or even override verbal communication. Deep multi-modal learning techniques have shown promise in understanding and analyzing these diverse aspects of BL. The survey emphasizes their applications to BL generation and recognition. Several common BLs are considered i.e., Sign Language (SL), Cued Speech (CS), Co-speech (CoS), and Talking Head (TH), and we have conducted an analysis and established the connections among these four BL for the first time. Their generation and recognition often involve multi-modal approaches. Benchmark datasets for BL research are well collected and organized, along with the evaluation of SOTA methods on these datasets. The survey highlights challenges such as limited labeled data, multi-modal learning, and the need for domain adaptation to generalize models to unseen speakers or languages. Future research directions are presented, including exploring self-supervised learning techniques, integrating contextual information from other modalities, and exploiting large-scale pre-trained multi-modal models. In summary, this survey paper provides a comprehensive understanding of deep multi-modal learning for various BL generations and recognitions for the first time. By analyzing advancements, challenges, and future directions, it serves as a valuable resource for researchers and practitioners in advancing this field. n addition, we maintain a continuously updated paper list for deep multi-modal learning for BL recognition and generation: https://github.com/wentaoL86/awesome-body-language.

cs.CV↗