mlsanityData to 7 Oct 2026

Token-Disentangled Latent Test-Time Scaling for Vision-Language Reasoning

arXiv 2609.35228 ↗First posted 28 Sep 2026cs.CV cs.AIcited by 0

On the map: LLM reasoning → Multimodal reasoning

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 36 of its 37 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

36 of 37 references.

Closest ancestors · 13

Same direction, newest first

  1. 2026Seeing with You: Perception-Reasoning Coevolution for Multimodal Reasoning30 Mar
  2. 2026Bridging Perception and Reasoning: Token Reweighting for RLVR in Multimodal LLMs26 Mar
  3. 2026Vision-aligned Latent Reasoning for Multi-modal Large Language Model4 Feb
  4. 2025Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space14 Dec
  5. 2025Limits and Gains of Test-Time Scaling in Vision-Language Reasoning11 Dec
  6. 2025Spotlight on Token Perception for Multimodal Reinforcement Learning10 Oct
  7. 2025Efficient Test-Time Scaling for Small Vision-Language Models3 Oct
  8. 2025InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency25 Aug
  9. 2025Visual-RFT: Visual Reinforcement Fine-Tuning3 Mar
  10. 2024LogicVista: Multimodal LLM Logical Reasoning Benchmark in Visual Contexts6 Jul
  11. 2024Visual CoT: Advancing Multi-Modal Language Models with a Comprehensive Dataset and Benchmark for Chain-of-Thought Reasoning25 Mar
  12. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct
1 more
  1. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering20 Sep

Other references · 23

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  3. 2025Qwen2.5-VL Technical Reportcited by 5,418
  4. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  5. 2025s1: Simple test-time scalingcited by 1,311
  6. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 946
  7. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 893
  8. 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 763
  9. 2024Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMscited by 719
  10. 2023HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Modelscited by 572
  11. 2025Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approachcited by 372
  12. 2025Reinforcement Learning with Verifiable Rewards Implicitly Incentivizes Correct Reasoning in Base LLMscited by 276
11 more
  1. 2022Generating Sequences by Learning to Self-Correctcited by 271
  2. 2025LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Trainingcited by 233
  3. 2024LLaVA-Critic: Learning to Evaluate Multimodal Modelscited by 132
  4. 2025SoftCoT: Soft Chain-of-Thought for Efficient Reasoning with LLMscited by 110
  5. 2025Seek in the Dark: Reasoning via Test-Time Instance-Level Policy Gradient in Latent Spacecited by 40
  6. 2025SoftCoT++: Test-Time Scaling with Soft Chain-of-Thought Reasoningcited by 38
  7. 2025What Kind of Visual Tokens Do We Need? Training-free Visual Token Pruning for Multi-modal Large Language Models from the Perspective of Graphcited by 35
  8. 2025AlphaOne: Reasoning Models Thinking Slow and Fast at Test Timecited by 27
  9. 2025Xiaomi MiMo-VL-Miloco Technical Reportcited by 15
  10. 2025Parallel Test-Time Scaling for Latent Reasoning Modelscited by 12
  11. 2025LatentEvolve: Self-Evolving Test-Time Scaling in Latent Spacecited by 8

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Self-Consistent Latent Reasoning: Long Latent Sequence Reasoning for Vision-Language Model
  2. 2026Leveraging Latent Visual Reasoning in Silence
  3. 2026Visual Enhanced Depth Scaling for Multimodal Latent Reasoning
  4. 2026Visual Latents Know More Than They Say: Unsilencing Latent Reasoning in MLLMs
  5. 2026Rethinking Latent Visual Reasoning: Grounding Latent Reasoning in Visual Evidence
  6. 2026What's Holding Back Latent Visual Reasoning?
  7. 2026Beyond Visual Memory: Mechanistic Diagnostics of Latent Visual Reasoning
  8. 2026On Test-Time Scaling for Vision-Language Models