arXiv subjects

Jia-Hong Huang

Publications and source records attributed to Jia-Hong Huang.

At least 19 recordsLinked to original sources

arXiv2026

SpeechGym: An Audio-Native Gym for Training Voice Agents via Reinforcement Learning

Voice agents must call tools and hold multi-turn dialogue entirely through speech, yet the dominant paradigm trains them in text. Existing frameworks either cascade TTS and ASR around a proprietary voice API, where gradients cannot flow and per-call cost makes on-policy reinforcement learning prohibitive, or stay in text: they measure voice agents but cannot improve them. We present SpeechGym, an audio-native agentic environment in which two omni-modal models converse in native audio, with no external ASR or TTS and no API boundary, over the unmodified tasks, tools and success check of an established text agentic benchmark, so that the interaction modality is the only variable and the loop stays local and trainable end to end. Audio agentic capability does not follow from audio understanding. The failures speech introduces are perceptual rather than reasoning deficits: the agent picks the right tool and the right argument slot but fills it with a value misheard from the waveform, and that single error cascades into a failed call, a retry of the same call, and a wasted step budget. A second failure is behavioural: under an insistent caller the agent performs an unauthorised write and ends the episode believing it helped. Both are trainable, because the environment labels them for free: a call with a misheard argument fails against the database while a correct one succeeds. The obstacle is sparsity, not signal. Outcome-only GRPO is gradient-starved here, since almost every rollout group fails identically, while a per-turn process reward crediting each successful tool call restores variance to nearly every group. Trained this way, the agent transfers with no further tuning to an independently implemented voice benchmark, more than doubling task success and carrying an open-weights model from last place to second on that leaderboard, while using fewer turns and tokens than before training.

cs.SD↗

arXiv2026

Can We Read the Mind of an Audio LLM? A Verbalizable, Multilingual Middle-Layer Workspace

An audio language model is a black box in a specific way: we see what it says, never what it works out on the way there, and chain-of-thought monitoring helps only if the model writes its reasoning down. Reading a base Qwen3-Omni with a logit lens at the audio-token positions, we find that the answer to a spoken question becomes legible - in words - in the model's middle layers, before it emits any token. Five findings follow. (1) The readout carries concepts in neither the question, the options, nor the model's own transcription: on a clip whose verbatim transcription is empty garbling, it reconstructs Watergate and scandal, passes through the role president, and resolves to Nixon - a hidden multi-hop chain, read with no chain-of-thought. (2) The content is language-agnostic: one audio-inferred concept surfaces in several scripts at once, and 38% of top-1 readouts are Chinese on English inputs. (3) It is paralinguistic: given the same clip as audio and as the model's own emotion-free caption, the audio mind forms the sound source, speaker role, or affect that the caption discards, and answers correctly more often. (4) The audio-driven signal is absent at the input, turns on about a tenth of the way into the network, separates most cleanly from the text prior in the middle band (35-80% of depth), and activation patching shows it is causally used and committed before the last fifth of the layers. (5) Deleting single layers maps the pipeline: reading the sound in is localized to the entry layers and answer delivery to the output layer, while retrieval is distributed across the interior. Throughout, a waveform-swap control - identical text, only the sound changed - isolates the audio-driven signal from a prior over the printed options. This is a qualitative account of what an audio model works out before it speaks: the quantities are controls, not benchmark scores.

cs.SD↗

arXiv2026

MMArt: A Multi-Perspective Multimodal Dataset for Visual Art Understanding

Recent vision-language models demonstrate impressive general visual understanding, yet their art interpretation remains shallow: they describe surface content but struggle with formal analysis, grounded historical interpretation, or affective characterization. We argue this is not only a model but also a dataset limitation. Existing art datasets are single perspective resources, where no dataset provides narrative, formal, emotional, and historical perspectives simultaneously for the same artworks. We introduce MMArt, a large-scale dataset of 74,234 WikiArt paintings, each annotated with four independently annotated perspectives plus a harmonized unified caption, produced by specialized vision-language models or human annotation and validated through complementary quality evaluations. Two complementarity analyses establish that perspectives encode genuinely distinct information. A generative analysis shows that formal analysis descriptions best preserve compositional style, and historical descriptions carry strong affective signal in reconstructed images. A discriminative retrieval analysis reveals task-asymmetry: narrative descriptions drive retrieval (R@1 = 44.0%), while formal descriptions, strongest for reconstruction, are nearly nondiscriminative at retrieval scale (R@1 = 7.8%). Leave-one-out analysis further confirms that historical descriptions are the least replaceable perspective across both tasks. Together, the two analyses establish that no single perspective suffices for all tasks, directly motivating MMArt multi-perspective design. The dataset, code, and additional information are available at https://shuaiwang97.github.io/MMArt/.

cs.CV↗

arXiv2026

Time Imprint: Learning Time-Aware Representations in Multi-Modal Knowledge Graphs

Multi-Modal Knowledge Graphs (MMKGs) enrich entities with multiple modalities such as text and images, yet entities with highly similar multi-modal features remain difficult to distinguish. Temporal information of an entity can serve as an additional modality to disambiguate such entities, but existing approaches rarely treat time as a separate modality alongside text and images due to two major challenges: (1) sparse temporal semantics, which hinder alignment with richer modalities, and (2) multiple timestamps, which introduce noise or reduce robustness in representation learning. To address these challenges, we propose Time Imprint, a framework that treats time as an entity-level modality and jointly aligns temporal, textual, and visual representations via a three-view contrastive objective. Additionally, to mitigate multi-timestamp ambiguity, Time Imprint studies a compact timestamp subset selection design space and aggregates the selected timestamps into a discriminative temporal embedding with attention pooling, balancing temporal specificity and robustness. Experiments on three MMKG benchmarks demonstrate that Time Imprint achieves state-of-the-art link prediction performance, improving Hits@1 by up to 6.07\% overall and yielding up to 58\% gains on the subset of the top-1\% ambiguity samples. We further examine different fusion strategies and the sensitivity to timestamp availability and quality, clarifying when and why time-as-modality is most beneficial, while adding only modest training overhead. We release our code at https://anonymous.4open.science/r/Time-Imprint.

cs.CV↗

arXiv2026

Keeping the Evidence Chain: Semantic Evidence Allocation for Training-Free Token Pruning in Video Temporal Grounding

Video Temporal Grounding (VTG) localizes the temporal boundaries of query-relevant moments in long, untrimmed videos, making video-language-model prohibitively expensive. While recent training-free token pruning has shown success in video question answering, naively applying these objectives to VTG causes drastic degradation, as VTG crucially depends on boundary-sensitive evidence and cross-frame reasoning chains. We therefore identify two VTG-specific pruning principles: evidence retention, which keeps query-critical patches especially around event boundaries, and connectivity strength, which preserves cross-frame connectivity for long-range evidence aggregation. Building on these insights, we propose SemVID, a training-free pruning framework that constructs a compact yet coherent token subset with complementary semantic roles. SemVID first allocates per-frame budgets by balancing query relevance and inter-frame variation to avoid over-pruned segments, and then selects three types of tokens: object tokens for diverse query-critical evidence, motion tokens to capture meaningful transitions and serve as cross-frame relays, and context tokens for scene continuity. Extensive experiments show that SemVID achieves a strong accuracy-efficiency trade-off, retaining up to 95.4% mIoU with only 12.5% visual tokens and delivering up to a 5.8x prefill speedup, consistently outperforming prior methods under the same budgets. Our code is available at https://github.com/JiaqiLi404/SemVID

cs.CV↗

arXiv2026

Spectral-Progressive Thought Flow for Lightweight Multimodal Reasoning

Multimodal spatial reasoning often relies on long chains of intermediate textual and visual thoughts, where accumulating visual tokens and dense cross-modal attention incur substantial computation and memory overhead. To address this challenge, we propose Spectral-Progressive Thought Flow (SpecFlow), a novel lightweight multimodal spatial reasoning framework that represents intermediate visual thoughts in a fixed-size discrete cosine space. By exploiting strong energy compaction, SpecFlow preserves global layout and relational structure while introducing high-frequency details only when increased spatial precision is required. To align visual state evolution with linguistic intent, classifier-free guidance enables autoregressive textual thoughts to steer flow-based updates of the visual workspace/state without expanding the context. As a result, SpecFlow maintains a bounded visual workspace whose updates depend only on the current visual state and accumulated textual trace, enabling long-horizon inference with stable latency and memory usage independent of reasoning depth. Empirical results show that SpecFlow achieves competitive or superior reasoning performance while reducing computation and KV cache costs by up to 2.1 times.

cs.LG↗

arXiv2026

A-MAR: Agent-based Multimodal Art Retrieval for Fine-Grained Artwork Understanding

Understanding artworks requires multi-step reasoning over visual content and cultural, historical, and stylistic context. While recent multimodal large language models show promise in artwork explanation, they rely on implicit reasoning and internalized knowl- edge, limiting interpretability and explicit evidence grounding. We propose A-MAR, an Agent-based Multimodal Art Retrieval framework that explicitly conditions retrieval on structured reasoning plans. Given an artwork and a user query, A-MAR first decomposes the task into a structured reasoning plan that specifies the goals and evidence requirements for each step. Retrieval is then conditionedon this plan, enabling targeted evidence selection and supporting step-wise, grounded explanations. To evaluate agent-based multi- modal reasoning within the art domain, we introduce ArtCoT-QA. This diagnostic benchmark features multi-step reasoning chains for diverse art-related queries, enabling a granular analysis that extends beyond simple final answer accuracy. Experiments on SemArt and Artpedia show that A-MAR consistently outperforms static, non planned retrieval and strong MLLM baselines in final explanation quality, while evaluations on ArtCoT-QA further demonstrate its advantages in evidence grounding and multi-step reasoning ability. These results highlight the importance of reasoning-conditioned retrieval for knowledge-intensive multimodal understanding and position A-MAR as a step toward interpretable, goal-driven AI systems, with particular relevance to cultural industries. The code and data are available at: https://github.com/ShuaiWang97/A-MAR.

cs.AI↗

arXiv2026

A Novel Automatic Framework for Speaker Drift Detection in Synthesized Speech

Recent diffusion-based text-to-speech (TTS) models achieve high naturalness and expressiveness, yet often suffer from speaker drift, a subtle, gradual shift in perceived speaker identity within a single utterance. This underexplored phenomenon undermines the coherence of synthetic speech, especially in long-form or interactive settings. We introduce the first automatic framework for detecting speaker drift by formulating it as a binary classification task over utterance-level speaker consistency. Our method computes cosine similarity across overlapping segments of synthesized speech and prompts large language models (LLMs) with structured representations to assess drift. We provide theoretical guarantees for cosine-based drift detection and demonstrate that speaker embeddings exhibit meaningful geometric clustering on the unit sphere. To support evaluation, we construct a high-quality synthetic benchmark with human-validated speaker drift annotations. Experiments with multiple state-of-the-art LLMs confirm the viability of this embedding-to-reasoning pipeline. Our work establishes speaker drift as a standalone research problem and bridges geometric signal analysis with LLM-based perceptual reasoning in modern TTS.

cs.SD↗

arXiv2026

Are a Thousand Words Better Than a Single Picture? Beyond Images -- A Framework for Multi-Modal Knowledge Graph Dataset Enrichment

Multi-Modal Knowledge Graphs (MMKGs) benefit from visual information, yet large-scale image collection is hard to curate and often excludes ambiguous but relevant visuals (e.g., logos, symbols, abstract scenes). We present Beyond Images, an automatic data-centric enrichment pipeline with optional human auditing. This pipeline operates in three stages: (1) large-scale retrieval of additional entity-related images, (2) conversion of all visual inputs into textual descriptions to ensure that ambiguous images contribute usable semantics rather than noise, and (3) fusion of multi-source descriptions using a large language model (LLM) to generate concise, entity-aligned summaries. These summaries replace or augment the text modality in standard MMKG models without changing their architectures or loss functions. Across three public MMKG datasets and multiple baseline models, we observe consistent gains (up to 7% Hits@1 overall). Furthermore, on a challenging subset of entities with visually ambiguous logos and symbols, converting images into text yields large improvements (201.35% MRR and 333.33% Hits@1). Additionally, we release a lightweight Text-Image Consistency Check Interface for optional targeted audits, improving description quality and dataset reliability. Our results show that scaling image coverage and converting ambiguous visuals into text is a practical path to stronger MMKG completion. Code, datasets, and supplementary materials are available at https://github.com/pengyu-zhang/Beyond-Images.

cs.CV↗

arXiv2025

Continuous-Token Diffusion for Speaker-Referenced TTS in Multimodal LLMs

Unified architectures in multimodal large language models (MLLM) have shown promise in handling diverse tasks within a single framework. In the text-to-speech (TTS) task, current MLLM-based approaches rely on discrete token representations, which disregard the inherently continuous nature of speech and can lead to loss of fine-grained acoustic information. In this work, we investigate the TTS within the MLLM paradigm using continuous speech representations. We design a dual-head architecture and implement two complementary training strategies for a robust model. (1) A diffusion head generating continuous speech representations is added on the MLLM, which is on frame-level and strictly autoregressive. (2) The original language model head is retained to preserve multitask capability and to control the start and end of speech synthesis. (3) Masked training is employed to address exposure bias in autoregressive decoding. (4) To stabilize optimization, we propose a two-stage scheme where the LM is frozen in the second stage, ensuring the diffusion head learns from a fixed input distribution. Evaluations on LibriSpeech(PC) test-clean show that our approach achieves state-of-the-art autoregressive performance, with a WER of 1.95%, speaker similarity of 0.54, and UTMOS of 4.00. The two-stage training yields a 46% relative WER reduction over the one-stage training baseline. These results highlight the effectiveness of combining autoregressive modeling with continuous-token diffusion, supported by a two-stage training procedure.

eess.AS↗

arXiv2025

DeepEyeNet: Generating Medical Report for Retinal Images

The increasing prevalence of retinal diseases poses a significant challenge to the healthcare system, as the demand for ophthalmologists surpasses the available workforce. This imbalance creates a bottleneck in diagnosis and treatment, potentially delaying critical care. Traditional methods of generating medical reports from retinal images rely on manual interpretation, which is time-consuming and prone to errors, further straining ophthalmologists' limited resources. This thesis investigates the potential of Artificial Intelligence (AI) to automate medical report generation for retinal images. AI can quickly analyze large volumes of image data, identifying subtle patterns essential for accurate diagnosis. By automating this process, AI systems can greatly enhance the efficiency of retinal disease diagnosis, reducing doctors' workloads and enabling them to focus on more complex cases. The proposed AI-based methods address key challenges in automated report generation: (1) A multi-modal deep learning approach captures interactions between textual keywords and retinal images, resulting in more comprehensive medical reports; (2) Improved methods for medical keyword representation enhance the system's ability to capture nuances in medical terminology; (3) Strategies to overcome RNN-based models' limitations, particularly in capturing long-range dependencies within medical descriptions; (4) Techniques to enhance the interpretability of the AI-based report generation system, fostering trust and acceptance in clinical practice. These methods are rigorously evaluated using various metrics and achieve state-of-the-art performance. This thesis demonstrates AI's potential to revolutionize retinal disease diagnosis by automating medical report generation, ultimately improving clinical efficiency, diagnostic accuracy, and patient care.

eess.IV↗

arXiv2025

MaCP: Minimal yet Mighty Adaptation via Hierarchical Cosine Projection

We present a new adaptation method MaCP, Minimal yet Mighty adaptive Cosine Projection, that achieves exceptional performance while requiring minimal parameters and memory for fine-tuning large foundation models. Its general idea is to exploit the superior energy compaction and decorrelation properties of cosine projection to improve both model efficiency and accuracy. Specifically, it projects the weight change from the low-rank adaptation into the discrete cosine space. Then, the weight change is partitioned over different levels of the discrete cosine spectrum, and each partition's most critical frequency components are selected. Extensive experiments demonstrate the effectiveness of MaCP across a wide range of single-modality tasks, including natural language understanding, natural language generation, text summarization, as well as multi-modality tasks such as image classification and video understanding. MaCP consistently delivers superior accuracy, significantly reduced computational complexity, and lower memory requirements compared to existing alternatives.

cs.LG↗

arXiv2025

MaCP: Minimal yet Mighty Adaptation via Hierarchical Cosine Projection

cs.LG↗

arXiv2025

SSH: Sparse Spectrum Adaptation via Discrete Hartley Transformation

Low-rank adaptation (LoRA) has been demonstrated effective in reducing the trainable parameter number when fine-tuning a large foundation model (LLM). However, it still encounters computational and memory challenges when scaling to larger models or addressing more complex task adaptation. In this work, we introduce Sparse Spectrum Adaptation via Discrete Hartley Transformation (SSH), a novel approach that significantly reduces the number of trainable parameters while enhancing model performance. It selects the most informative spectral components across all layers, under the guidance of the initial weights after a discrete Hartley transformation (DHT). The lightweight inverse DHT then projects the spectrum back into the spatial domain for updates. Extensive experiments across both single-modality tasks such as language understanding and generation and multi-modality tasks such as video-text understanding demonstrate that SSH outperforms existing parameter-efficient fine-tuning (PEFT) methods while achieving substantial reductions in computational cost and memory requirements.

cs.CV↗

arXiv2025

Gradient Weight-normalized Low-rank Projection for Efficient LLM Training

Large Language Models (LLMs) have shown remarkable performance across various tasks, but the escalating demands on computational resources pose significant challenges, particularly in the extensive utilization of full fine-tuning for downstream tasks. To address this, parameter-efficient fine-tuning (PEFT) methods have been developed, but they often underperform compared to full fine-tuning and struggle with memory efficiency. In this work, we introduce Gradient Weight-Normalized Low-Rank Projection (GradNormLoRP), a novel approach that enhances both parameter and memory efficiency while maintaining comparable performance to full fine-tuning. GradNormLoRP normalizes the weight matrix to improve gradient conditioning, facilitating better convergence during optimization. Additionally, it applies low-rank approximations to the weight and gradient matrices, significantly reducing memory usage during training. Extensive experiments demonstrate that our 8-bit GradNormLoRP reduces optimizer memory usage by up to 89.5% and enables the pre-training of large LLMs, such as LLaMA 7B, on consumer-level GPUs like the NVIDIA RTX 4090, without additional inference costs. Moreover, GradNormLoRP outperforms existing low-rank methods in fine-tuning tasks. For instance, when fine-tuning the RoBERTa model on all GLUE tasks with a rank of 8, GradNormLoRP achieves an average score of 80.65, surpassing LoRA's score of 79.23. These results underscore GradNormLoRP as a promising alternative for efficient LLM pre-training and fine-tuning. Source code: https://github.com/Jhhuangkay/Gradient-Weight-normalized-Low-rank-Projection-for-Efficient-LLM-Training

cs.LG↗

arXiv2024

Image2Text2Image: A Novel Framework for Label-Free Evaluation of Image-to-Text Generation with Text-to-Image Diffusion Models

Evaluating the quality of automatically generated image descriptions is a complex task that requires metrics capturing various dimensions, such as grammaticality, coverage, accuracy, and truthfulness. Although human evaluation provides valuable insights, its cost and time-consuming nature pose limitations. Existing automated metrics like BLEU, ROUGE, METEOR, and CIDEr attempt to fill this gap, but they often exhibit weak correlations with human judgment. To address this challenge, we propose a novel evaluation framework called Image2Text2Image, which leverages diffusion models, such as Stable Diffusion or DALL-E, for text-to-image generation. In the Image2Text2Image framework, an input image is first processed by a selected image captioning model, chosen for evaluation, to generate a textual description. Using this generated description, a diffusion model then creates a new image. By comparing features extracted from the original and generated images, we measure their similarity using a designated similarity metric. A high similarity score suggests that the model has produced a faithful textual description, while a low score highlights discrepancies, revealing potential weaknesses in the model's performance. Notably, our framework does not rely on human-annotated reference captions, making it a valuable tool for assessing image captioning models. Extensive experiments and human evaluations validate the efficacy of our proposed Image2Text2Image evaluation framework. The code and dataset will be published to support further research in the community.

cs.CV↗

arXiv2024

Personalized Video Summarization using Text-Based Queries and Conditional Modeling

The proliferation of video content on platforms like YouTube and Vimeo presents significant challenges in efficiently locating relevant information. Automatic video summarization aims to address this by extracting and presenting key content in a condensed form. This thesis explores enhancing video summarization by integrating text-based queries and conditional modeling to tailor summaries to user needs. Traditional methods often produce fixed summaries that may not align with individual requirements. To overcome this, we propose a multi-modal deep learning approach that incorporates both textual queries and visual information, fusing them at different levels of the model architecture. Evaluation metrics such as accuracy and F1-score assess the quality of the generated summaries. The thesis also investigates improving text-based query representations using contextualized word embeddings and specialized attention networks. This enhances the semantic understanding of queries, leading to better video summaries. To emulate human-like summarization, which accounts for both visual coherence and abstract factors like storyline consistency, we introduce a conditional modeling approach. This method uses multiple random variables and joint distributions to capture key summarization components, resulting in more human-like and explainable summaries. Addressing data scarcity in fully supervised learning, the thesis proposes a segment-level pseudo-labeling approach. This self-supervised method generates additional data, improving model performance even with limited human-labeled datasets. In summary, this research aims to enhance automatic video summarization by incorporating text-based queries, improving query representations, introducing conditional modeling, and addressing data scarcity, thereby creating more effective and personalized video summaries.

cs.CV↗

arXiv2024

QFMTS: Generating Query-Focused Summaries over Multi-Table Inputs

Table summarization is a crucial task aimed at condensing information from tabular data into concise and comprehensible textual summaries. However, existing approaches often fall short of adequately meeting users' information and quality requirements and tend to overlook the complexities of real-world queries. In this paper, we propose a novel method to address these limitations by introducing query-focused multi-table summarization. Our approach, which comprises a table serialization module, a summarization controller, and a large language model (LLM), utilizes textual queries and multiple tables to generate query-dependent table summaries tailored to users' information needs. To facilitate research in this area, we present a comprehensive dataset specifically tailored for this task, consisting of 4909 query-summary pairs, each associated with multiple tables. Through extensive experiments using our curated dataset, we demonstrate the effectiveness of our proposed method compared to baseline approaches. Our findings offer insights into the challenges of complex table reasoning for precise summarization, contributing to the advancement of research in query-focused multi-table summarization.

cs.CL↗