Searcharxiv⌕ Search

arXiv subjects

Junmo Kim

Publications and source records attributed to Junmo Kim.

At least 19 recordsLinked to original sources
arXiv2026

AccentCanvas: Activity-Guided Bidirectional Accent Conversion

Accent conversion for dubbing must change pronunciation while preserving content, voice, and source timing. Normalization alone cannot produce requested non-native accents, and matching total duration does not ensure local speech/pause alignment. We present AccentCanvas, a shared discrete diffusion model for normalization and L1-to-L2 conversion, trained on cross-paired shared-prompt recordings and synthetic pairs. Frame-level speech activity and repetition-preserving units guide local timing through synthesis. Across both tasks, AccentCanvas achieves the highest accent conversion performance (Accent Arrival) and mean listener-rated accent similarity among compared systems, with over 73% lower speech-boundary error than the best baseline for each task. These gains accompany higher L1-to-L2 WER than TokAN and Vevo-Style, highlighting a tradeoff between requested-accent resemblance and ASR-based content recognition. https://smokedindia.github.io/accentcanvas/

eess.AS↗
arXiv2026

HyperVAttention: Efficient Sparse Attention with Spatio-Temporal Clustering for Video Diffusion

Video Diffusion Transformers (VDiTs) have demonstrated significant capabilities in high-fidelity video generation. However, their ability to produce long-duration videos is fundamentally constrained by the quadratic complexity of the self-attention mechanism. Recent clustering-based sparse attention methods improve the quality-speed trade-off by grouping semantically similar tokens, but their practical efficiency remains limited by two bottlenecks: substantial clustering overhead and low CTA utilization caused by irregular cluster-induced blocks. We propose HyperVAttention (HVA), a training-free sparse attention framework that addresses both bottlenecks jointly. To reduce clustering overhead, we introduce 3D local-window clustering, which exploits the spatio-temporal locality of video tokens to restrict centroid search to fixed local neighborhoods, and implement it with a custom Triton kernel for efficient execution. We further propose a hybrid clustering strategy that performs full clustering only at anchor steps and updates only subset tokens at intermediate steps, leveraging the temporal stability of cluster assignments across denoising steps. To improve CTA utilization, we present hardware-aware cluster merging that minimizes CTA-aligned execution cost through parallel agglomerative merging, improving block density and approximation fidelity by utilizing idle tile capacity. Together, these components reduce clustering overhead, avoid redundant updates, and better align sparse attention with the fixed tile structure of modern GPU kernels. Experiments on video generation show that HVA advances the Pareto frontier of training free sparse attention, achieving higher fidelity than SVG2 at comparable latency and a $2.13\times$ speedup over full attention on HunyuanVideo, or $2.23\times$ with its faster variant HVA-Flash.

cs.CV↗
arXiv2026

Imitation Dominates Reinforcement: Direct In-Context RL Is Closer to ICL Than RL

LLM agents increasingly improve at inference time by accumulating experience in context rather than by updating parameters. This process is often described as in-context reinforcement learning (ICRL). Whether in-context learning (ICL) can actually play the role of RL, however, has not been tested. We study this question in its simplest form, direct ICRL, where the model conditions directly on raw trajectory-reward pairs, and ask whether the reward acts as a learning signal. Through controlled experiments on four benchmarks across six models, we find that the reward is read, but its effect is small: flipping, randomizing, or removing the reward leaves the improvement curve almost unchanged, and this holds even under meta-prompts that explicitly instruct the model to explore, exploit, or reason over rewards. Trajectories drive improvement, but not through their semantic content: shuffled or corrupted trajectories work as well as real ones. These patterns closely mirror those known in ICL, suggesting that direct ICRL is better understood as a special case of ICL than as inference-time RL. This reframing has implications for agent memory design: ICL factors such as input distribution and demonstrations may matter more than RL elements such as reward shaping and exploration.

cs.LG↗
arXiv2026

Continuous Audio Thinking for Large Audio Language Models

Large audio language models (LALMs) have shown impressive capabilities on diverse audio understanding tasks, ranging from speech transcription to music analysis. However, because LALMs are typically trained to produce text-aligned responses, their hidden states are progressively shaped for text generation rather than for preserving acoustic information. As a result, the diverse acoustic content that audio carries, such as phonetic detail, prosody, sound events, affect, and pitch, is lost along the way and difficult to leverage in the response. We introduce Continuous Audio Thinking (CoAT), a framework that equips audio language models with a continuous latent workspace for organizing acoustic information prior to response generation, grounded by distillation from audio experts. Within the thinking space, the model can utilize the rich acoustic information provided by expert distillation when generating its response. Furthermore, the proposed continuous thinking block can be processed in a single prefill, so CoAT does not require additional autoregressive decoding cost over the baseline. Across three LALMs, Qwen2-Audio, Qwen2.5-Omni-7B, and Audio Flamingo~3, performance gains on a broad benchmark suite spanning audio reasoning, audio understanding, music classification, speech emotion, and speech transcription demonstrate the effectiveness of CoAT. Further analysis confirms that the auxiliary supervision propagates from the thinking positions to the model's textual responses.

cs.CL↗
arXiv2026

VGGT-Bridge: Beyond Sequential Pose Graphs via Coarse-Stride Skip Edges

Feed-forward visual geometry transformers such as VGGT reconstruct dense 3D structure from images in a single forward pass, simplifying multi-view 3D reconstruction. However, their quadratic attention complexity makes them difficult to scale to long sequences with thousands of frames. Chunk-and-align frameworks address this by splitting a long sequence into overlapping chunks and stitching their local reconstructions into a pose graph. Yet existing methods connect only sequentially adjacent chunks, so small per-frame errors accumulate along the chain into large-scale drift. To move beyond sequential edges, we propose VGGT-Bridge, which adds long-range skip edges that directly constrain non-adjacent chunks without retraining. By running VGGT on sparsely sampled coarse chunks, each coarse chunk bridges distant fine chunks into a single direct constraint. We further turn VGGT's first-frame scale bias into a drift correction by feeding selected coarse chunks in reverse, and a loop-aware policy keeps this reversal compatible with existing loop closures. VGGT-Bridge reduces ATE by 28.3% on KITTI Odometry, 18.8% on Virtual KITTI, and 10.0% on Waymo Open over the SwiftVGGT baseline, achieving the best performance among all chunk-and-align methods. Our code is available at https://github.com/BeCow5X5/VGGT-Bridge.

cs.CV↗
arXiv2026

SCALE: Improving Guidance-Free Visual Generation via Self-Contrastive Alignment for Likelihood Estimation

Classifier-free guidance (CFG) is essential for improving sample quality in visual generation, but it incurs substantial sampling overhead. Existing guidance-free approaches halve this cost by approximating the CFG with a single model inference, but their performance is bounded by the CFG and inherits its limitations. In this paper, we propose self-contrastive guidance (SCG), which contrasts the model's prediction on real and self-generated samples to obtain a direct, model-based signal that pushes generation toward the true data manifold. Integrating SCG into the standard likelihood objective via reparameterization yields Self-Contrastive Alignment for Likelihood Estimation (SCALE), a guidance-free training framework that produces SCG-enhanced predictions in a single forward pass. Across class-conditional and text-to-image generation on diffusion and autoregressive models, SCALE consistently outperforms CFG-mimicking guidance-free baselines and improves further when combined with DPO, demonstrating that the self-contrastive signal yields gains complementary to both CFG and preference-based fine-tuning.

cs.CV↗
arXiv2026

Calibration-Free Early Stopping for Chain-of-Thought Reasoning via E-processes

Reasoning large language models (LLMs) benefit from test-time scaling, motivating adaptive policies that spend tokens when they help and stop early otherwise. Existing rules halt when a confidence or stability signal crosses a threshold, but such thresholds rarely transfer across models and datasets, and validation-calibrated rules lose their guarantee under shifts in test-time horizon or signal dynamics. We reframe adaptive stopping as sequential evidence accumulation along the reasoning trajectory rather than threshold calibration on held-out data. Our framework runs two e-processes on a shared per-chunk signal: a commit stream tracks the signal's position relative to its slow trend, and an abstain stream tracks its signed velocity, making the two anti-correlated by construction. Under signal-level nulls on the per-chunk increments, Ville's inequality yields an anytime-valid signal-level false-alarm bound at user-specified significance levels, with no held-out calibration. Across various model--dataset combinations, the framework matches or exceeds calibration-based stopping on the token--accuracy Pareto front using zero calibration data, and the same significance level transfers unchanged across cells where calibration becomes infeasible.

cs.CL↗
arXiv2026

Syn-Omni: Structured Specialization and Progressive Collaboration for Omnimodal Embeddings

Omnimodal embeddings naturally involve both shared representations and modality-specific features across heterogeneous inputs. However, existing omnimodal embedding methods often rely on a single shared parameter space over mixed-modality data, limiting structural separation between universal and modality-specific representations. To address this, we propose Syn-Omni, a unified framework for structured omnimodal adaptation with modality specialization and controlled cross-modal collaboration. Specifically, we introduce Orthogonal Modality-Expert LoRA (OME-LoRA), which decomposes adaptation into a shared LoRA path for universal semantics and modality-expert LoRA paths for modality-aware specialization. Furthermore, Progressive Synergy Routing (PSR) enables experts to first establish modality-specific priors, then gradually interact with other modality-experts for cross-modal synergy. Evaluated across 81 diverse tasks spanning image, video, audio, and audiovisual modalities, Syn-Omni consistently outperforms omnimodal baselines, demonstrating the effectiveness of structured specialization and cross-modal progressive collaboration.

cs.CV↗
arXiv2026

Before It Fades: Reinforcing Temporal Representations at Inference Time in VideoLLMs

Video Large Language Models (VideoLLMs) receive frames in sequential order and interpret how visual content evolves along the temporal axis, yet temporal reasoning remains a persistent weakness across architectures. Reversing the frame order of a video, a transformation that should invert temporal answers, often leaves the final prediction unchanged. We investigate where this failure originates by defining the temporal divergence vector $τ_l$, the layer-wise representational difference induced by reversing temporal order. Tracking its magnitude across layers reveals a consistent temporal divergence profile where the divergence peaks at intermediate layers and progressively diminishes toward the output. We confirm this peak is specific to temporal reasoning and functionally critical for predictions, establishing that VideoLLMs acquire temporal information at intermediate layers but fail to maintain it to the output. This progressive fading motivates our method, Temporal Activation Injection (TAI), which extracts $τ_l$ at the peak of the profile for each input and reinjects it into subsequent layers following the measured decay. TAI requires no training and consistently improves temporal reasoning across three VideoLLMs and four benchmarks with negligible impact on non-temporal tasks. Code is available at https://github.com/Youngwoo-git/Before-It-Fades.

cs.CV↗
arXiv2026

PePESeg3D: Perception Prior Enhances Multi-Scale Segmentation for 3D Gaussian Splatting

Recent advancements in 3D Gaussian Splatting (3DGS) have extended its capabilities to multi-scale segmentation. Existing methods reconstruct a scene with Gaussian primitives and learn multi-scale segmentation features separately, which leaves the geometry unaware of semantic structure and the feature learning dependent on incomplete mask supervision. To address these limitations, we present PePESeg3D, a novel framework that injects perception priors into a multi-scale 3D Gaussian segmentation pipeline. To fully exploit perception priors, we integrate them not only into contrastive feature learning but also into the upstream geometry reconstruction. Specifically, PePE Reconstruction incorporates monocular depth and mask constraints to ensure semantically coherent object structures. Building on this aligned geometry, PePE Contrastive Learning leverages dense depth-color cues and view-consistent centroid supervision to compensate for the incompleteness of multi-scale masks obtained from a 2D foundation model. Extensive experiments on the SPIn-NeRF, LERF-Mask, and NVOS benchmarks demonstrate that PePESeg3D achieves state-of-the-art performance in both multi-scale segmentation and scene reconstruction, highlighting the importance of integrating perception priors into both geometry optimization and feature learning for accurate multi-scale 3D segmentation. Our code is available at https://github.com/BeCow5X5/PePESeg3D.

cs.CV↗
arXiv2026

Info3R: Information-Adaptive Test-Time Training for 3D Reconstruction

Transformer-based models have recently achieved strong performance on 3D reconstruction from images, and recent works extend them to process video streams in an online manner for real-world deployment. However, existing methods overlook two key signals when handling long image streams: the importance of each incoming frame and the information saturation of the model's internal state. In this paper, we propose Info3R, a novel information-adaptive test-time training method for the online 3D reconstruction. We introduce an information-aware state update that modulates the state update strength based on the redundancy and informativeness of each incoming frame. To restore the state's plasticity -- its capacity to incorporate new observations -- we propose a dynamic state reset, triggered by the cumulative magnitude of state updates and the model's prediction confidence and accompanied by an anchor-to-world alignment. Our method achieves consistent improvements on camera pose estimation, video depth estimation, and 3D reconstruction, while substantially mitigating the performance degradation in the long sequence evaluation. Notably, on KITTI Odometry, our method achieves on average 1.68x lower ATE than LongStream, demonstrating its robustness on extended outdoor sequences.

cs.CV↗
arXiv2026

Direct Diversity Optimization for Diverse Successful Trajectories in Preference Post-Training

LLM agents for sequential decision tasks are often post-trained with trajectory-level outcome labels, but such labels provide little supervision for preserving multiple successful branches from the same decision state. We study this problem as successful strategy coverage: how broadly a model realizes distinct successful strategies under a fixed rollout budget. We present Direct Diversity Optimization (DDO), an offline post-training method that combines Divergence-Tree Collection (DTC) with the Reference-Relative Target-Odds Objective (RTO). DTC constructs state-aligned branch sets rooted at shared decision states, and RTO trains the model to match reference-relative targets over successful alternatives. DDO achieves the strongest task success and successful strategy coverage among the compared post-training methods across BabyAI, BabaIsAI, and WebShop. It also achieves the highest recovery rate after local action replacement and higher task success and coverage than successful-only imitation and decoding-time diversification controls.

cs.CL↗
arXiv2026

ARGOS: Who, Where, and When in Agentic Multi-Camera Person Search

Existing person search methods assume access to complete visual queries or exhaustive tracking, yet real-world witness accounts are vague, partial, and spread across cameras and time. We introduce ARGOS (Agentic Retrieval with Grounded Observational Search), a benchmark and agent framework that recasts multi-camera person search from one-shot retrieval on a complete query into interactive reasoning from partial clues. To our knowledge, ARGOS is the first interactive benchmark to couple witness dialogue with camera-network topology, requiring an agent to plan, question, and eliminate under information asymmetry. An ARGOS agent receives a vague witness statement and must decide what to ask, when to invoke spatial or temporal tools, and how to interpret ambiguous natural-language responses, all within a limited turn budget. To ground reasoning in physical constraints, the agent accesses a Spatio-Temporal Topology Graph (STTG) encoding camera connectivity and empirically validated transition times. The benchmark comprises 2{,}691 tasks across 14 real-world scenarios in three progressive tracks: semantic perception (\emph{Who}, 989 tasks), spatial reasoning (\emph{Where}, 550 tasks), and temporal reasoning (\emph{When}, 1{,}152 tasks). We propose Turn-Weighted Success (TWS) as the primary metric, jointly measuring correctness and turn efficiency. Experiments with four LLM backbones show the benchmark is far from solved: the best agent achieves TWS of 0.383 (Track~2) and 0.590 (Track~3). Ablations confirm each component is essential: removing domain-specific tools drops Top-1 accuracy by up to 49.6 percentage points, and removing strategic reasoning halves TWS while barely affecting Top-1.

cs.CV↗
arXiv2026

Doesn't Stop Reasoning: Analysis of Spurious CoT Termination

Chain-of-thought (CoT) reasoning improves large reasoning models (LRMs) on complex tasks but often produces long, redundant traces. Recent training-free early-exit methods shorten these traces by choosing an intermediate point to stop reasoning. We study one such strategy that injects an end-of-think token (EoT, ) at this point to trigger the reasoning-to-answering transition, and find that the injected EoT does not always induce a clean answering phase. Answering-phase generation can continue before the model regenerates another EoT, with the span preceding this regenerated EoT scaling with the reasoning tokens saved by early exit and exhibiting continued reasoning behavior. We call this spurious CoT termination, where reasoning-like generation continues into the answering phase. We hypothesize that insufficient attention to the injected EoT contributes to spurious CoT termination and probe this hypothesis with Exit-token Attention Biasing (EAB). Across four LRMs, five benchmarks, and two early-exit methods, increasing attention to the injected EoT reduces spurious CoT termination and answering-phase length. These results reveal a limitation of controlling LRMs by externally matching their explicit think-block format. Inserting the EoT token conforms to this format but does not by itself guarantee the intended reasoning-to-answering transition. Our code is available at https://github.com/Seunghee-Koh/Spurious-CoT-Termination.

cs.CL↗
arXiv2026

IWP: Token Pruning as Implicit Weight Pruning in Large Vision Language Models

Large Vision Language Models show impressive performance across image and video understanding tasks, yet their computational cost grows rapidly with the number of visual tokens. Existing token pruning methods mitigate this issue through empirical approaches while overlooking the internal mechanism of attention. In this paper, we propose a novel training free token pruning framework grounded in the dual form perspective of attention. We reformulate attention as an implicit linear layer whose weight matrix is the sum of rank 1 outer products, each generated by a single token's key value pair. Token pruning thus reduces to selecting an optimal subset of these rank 1 updates that best approximates the original dual weight matrix. Extending this perspective to standard softmax attention in LVLMs, we derive a novel metric quantifying both a token's information magnitude and information duplication. To efficiently select the subset with the proposed metric, we introduce Progressive Chunked Maximal Marginal Relevance. Extensive experiments demonstrate that our method achieves a better trade off between performance and efficiency, while providing another perspective on existing pruning approaches.

cs.CV↗
arXiv2026

MLLMCLIP: Feature-Level Distillation of MLLM for Robust Vision-Language Representations

Pretrained vision-language models such as CLIP excel at zero-shot recognition but often fail at compositionality, particularly attribute-object and relational structures. Recent studies mitigate this issue by augmenting training with synthetic hard negatives generated by a cascade of large language models and text-to-image models, which incurs substantial pipeline overhead. We instead propose MLLMCLIP, a heterogeneous distillation framework that transfers multimodal knowledge directly from a generative Multimodal Large Language Model (MLLM) teacher into a discriminative CLIP student, bypassing synthetic data entirely. To bridge the architectural mismatch between the two paradigms, we introduce an attention-based per-layer token selection and a CKA-based distillation loss. Compared to prior CLIP-enhancement methods, MLLMCLIP achieves state-of-the-art compositional accuracy while delivering consistent gains on standard zero-shot classification and image-text retrieval, showing that feature-level distillation strengthens both compositional and general vision-language representation capability.

cs.CV↗
arXiv2026

Stable-GFlowNet: Toward Diverse and Robust LLM Red-Teaming via Contrastive Trajectory Balance

Large Language Model (LLM) Red-Teaming, which proactively identifies vulnerabilities of LLMs, is an essential process for ensuring safety. Finding effective and diverse attacks in red-teaming is important, but achieving both is challenging. Generative Flow Networks (GFNs) that perform distribution matching are promising methods, but they are notorious for training instability and mode collapse. In particular, unstable rewards in red-teaming accelerate mode collapse. We propose Stable-GFN (S-GFN), which eliminates partition function $Z$ estimation in GFN and reduces training instability. S-GFN avoids $Z$ estimation through pairwise comparisons and employs a robust masking methodology against noisy rewards. Additionally, we propose a fluency stabilizer to prevent the model from getting stuck in local optima that produce gibberish. S-GFN provides more stable training while maintaining the optimal policy of GFN. We demonstrate the overwhelming attack performance and diversity of S-GFN across various settings. Our code can be found in https://github.com/kmc0207/Stable-GFN.

cs.LG↗
arXiv2026

Cross-Axis Feature Fusion with Joint-Wise Motion Difference Prediction for Text-Based 3D Human Motion Editing

We address text-based 3D human motion editing, where the goal is to preserve the style and structure of a source motion while applying edits described in natural language. The release of the MotionFix dataset has spurred active research into training-based diffusion models that directly generate an edited motion from a source motion and a text instruction. While previous works have focused primarily on learning when an edit should occur temporally, our goal is to create a model that understands not only this temporal aspect but also which specific joints are responsible for the change. Targeting this, we propose a novel architecture and a complementary auxiliary task to aid its training. Our architecture consists of two axis-anchored transformers, which extract distinct features along the joint and time dimensions respectively, and a cross-axis fusion block that integrates these representations. We further introduce an auxiliary task that trains the joint-anchored transformer to regress the Soft-DTW distance between source and target joint rotations. This objective teaches the module to understand which joints to modify and which to preserve. Through comprehensive experiments on the MotionFix dataset, we demonstrate that our method significantly improves semantic alignment with both the text instruction and the source motion, as well as the overall fidelity of the generated motion, achieving state-of-the-art results.

cs.CV↗