Sep 2026· Conference on Empirical Methods in Natural Language Processing· Vol abs/2609.35228· 0 citations· 37 references
Computer Science
TL;DR
This work presents Token-Disentangled Latent Test-Time Scaling, an inference-time framework that makes latent refinement token-role-aware and outperforms strong output-space test-time scaling baselines under matched decoded-candidate budgets.
Abstract
Latent test-time scaling improves reasoning by refining hidden states during inference, but existing methods typically apply a single scalar reward to all editable latent tokens. For multimodal large language models, this global update ignores that generated tokens play different roles: some are sensitive to visual evidence, while others correspond to uncertain reasoning decisions. We present Token-Disentangled Latent Test-Time Scaling, an inference-time framework that makes latent refinement token-role-aware. Starting from an initial generated trajectory, we optimize a short hidden-state prefix while routing perception-side visual feedback to image-sensitive tokens and reasoning feedback to high-entropy tokens. Tokens selected by neither route are constrained by an anchor regularizer. Across both perception and reasoning benchmarks on Qwen2.5-VL-7B and InternVL3.5-8B, our method lifts macro accuracy over CoT by +2.57 and +1.51 respectively, and outperforms strong output-space test-time scaling baselines under matched decoded-candidate budgets. Code is available at https://github.com/Qwen-Applications/TD-LTTS.
ReaLVR is proposed, which brings visual-evidence supervision to the model's own free-running latent trajectories, and is the first to scale visual reasoning in latent space, showing that the framework continues to deliver robust improvements at frontier model scales up to 235B.
Xi Xiao, Tian-Chen Zhao, Youngeun Kim et al.· arXiv.org· 2 citations
Decision-Flow Sampling (DF-Sample), a training-free, data-free inference-time framework that constructs a hierarchical reasoning tree, scores terminal nodes for quality, and back-propagates utilities to inform each intermediate branching decision, indicating substantial latent reasoning potential in pretrained base mod...
Vision-Language-Action (VLA) models remain brittle under visual distribution shifts, often relying on spurious correlations tied to domain-specific factors rather than task-relevant structure. We propose Domain-Invariant Latent Lookahead (DILL), a representation-learning framework that mitigates shortcut learning in VL...
Junghyun Kim, Ngseo Kim, Chung-Woo Lee et al.· 0 citations
Latent-OPD is proposed, which augments OPD with trajectory-level latent distillation and introduces a progressive teacher-lookahead strategy, which aligns middle-to-late student layers with increasingly deeper teacher layers, establishing Latent-OPD as a highly effective approach to frame-efficient video reasoning.
Aoni Shen, Yongheng Zhang, Ying-Hui Li et al.· 2 citations
TTRSD separates update direction from update position, determined by group-relative advantages, from update position, determined by visual sensitivity, without requiring ground-truth labels, external verifiers, or a separate teacher, demonstrating cross-dataset generalization while preserving inherent reasoning integri...
LIRSeg is proposed, which fully replaces explicit CoT with a compact set of learnable latent tokens for reasoning segmentation, and introduces three complementary mechanisms from an information perspective: extreme-advantage sampling for selecting informative training signals, decoupled exploration-stability updates fo...
Known for his clear and elegant writing style, Bertsekas shaped fields from control and optimization to large-scale computation and artificial intelligence.