# Chain-of-Thought Compression Should Not Be Blind: V-Skip for Efficient Multimodal Reasoning via Dual-Path Anchoring

arXiv: https://arxiv.org/abs/2601.13879 · HTML: https://mlsanity.com/paper/2601.13879/

## Key facts

- Cited by 21 papers in the map (data through 14 Oct 2026).
- Placed in Multimodal reasoning, one of 17 directions in LLM reasoning.
- Builds on 17 of its 34 references that are in the map.
- First posted on arXiv 20 Jan 2026.

## Placed in

- LLM reasoning / [Multimodal reasoning](https://mlsanity.com/lr/multimodal-reasoning/)
- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 17 of 17)

- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [In-context Learning and Induction Heads](https://mlsanity.com/paper/2209.11895/) (2022)
- [Deep Variational Information Bottleneck](https://mlsanity.com/paper/1612.00410/) (2016)
- [Multimodal Chain-of-Thought Reasoning in Language Models](https://mlsanity.com/paper/2302.00923/) (2023)
- [Fast Transformer Decoding: One Write-Head is All You Need](https://mlsanity.com/paper/1911.02150/) (2019)
- [Object Hallucination in Image Captioning](https://mlsanity.com/paper/1809.02156/) (2018)
- [LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models](https://mlsanity.com/paper/2407.07895/) (2024)
- [LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression](https://mlsanity.com/paper/2403.12968/) (2024)
- [LLMLingua: Compressing Prompts for Accelerated Inference of Large Language Models](https://mlsanity.com/paper/2310.05736/) (2023)
- [TokenSkip: Controllable Chain-of-Thought Compression in LLMs](https://mlsanity.com/paper/2502.12067/) (2025)
- [Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey](https://mlsanity.com/paper/2503.12605/) (2025)
- [Learning to Compress Prompt in Natural Language Formats](https://mlsanity.com/paper/2402.18700/) (2024)
- [Token-Importance Guided Direct Preference Optimization](https://mlsanity.com/paper/2505.19653/) (2025)
- [Not All Errors Are Created Equal: ASCoT Addresses Late-Stage Fragility in Efficient LLM Reasoning](https://mlsanity.com/paper/2508.05282/) (2025)
- [Breaking the MoE LLM Trilemma: Dynamic Expert Clustering with Structured Compression](https://mlsanity.com/paper/2510.02345/) (2025)

## Built on by (most cited first, 21 of 21)

- [PointCoT: A Multi-modal Benchmark for Explicit 3D Geometric Reasoning](https://mlsanity.com/paper/2602.23945/) (2026)
- [Not All Queries Need Deep Thought: CoFiCot for Adaptive Coarse-to-fine Stateful Refinement](https://mlsanity.com/paper/2603.08251/) (2026)
- [PointRFT: Explicit Reinforcement Fine-tuning for Point Cloud Few-shot Learning](https://mlsanity.com/paper/2603.23957/) (2026)
- [CMHANet: A Cross-Modal Hybrid Attention Network for Point Cloud Registration](https://mlsanity.com/paper/2603.12721/) (2026)
- [IGASA: Integrated Geometry-Aware and Skip-Attention Modules for Enhanced Point Cloud Registration](https://mlsanity.com/paper/2603.12719/) (2026)
- [Pansharpening for Thin-Cloud Contaminated Remote Sensing Images: A Unified Framework and Benchmark Dataset](https://mlsanity.com/paper/2603.14952/) (2026)
- [FADE: Mitigating Hallucinations by Reducing Language-Prior Dominance in Large Vision-Language Models](https://mlsanity.com/paper/2606.29431/) (2026)
- [Parameter-Efficient Fine-Tuning for Spiking Point Cloud Models](https://mlsanity.com/paper/2607.29048/) (2026)
- [Robust Regularized Policy Iteration under Transition Uncertainty](https://mlsanity.com/paper/2603.09344/) (2026)
- [CFMS: A Coarse-to-Fine Multimodal Synthesis Framework for Enhanced Tabular Reasoning](https://mlsanity.com/paper/2604.10973/) (2026)
- [SeeMe: Mitigating Hallucinations in Large Vision-Language Models through Effective Visual Token Engineering](https://mlsanity.com/paper/2607.04163/) (2026)
- [Better Starts, Better Ends: Bootstrapped Iterative Self-Reasoning Distillation for Compressed Reasoning](https://mlsanity.com/paper/2607.15736/) (2026)
- [Can LLM-Generated Text Empower Surgical Vision-Language Pre-training?](https://mlsanity.com/paper/2604.18134/) (2026)
- [SF-AMS: Strategic Forgetting for Structured Memory in LLM Agent](https://mlsanity.com/paper/2607.22562/) (2026)
- [Bridging Efficiency and Transparency: Explainable CoT Compression in Multimodal Large Reasoning Models](https://mlsanity.com/paper/2602.09485/) (2026)
- [Controllable Singing Style Conversion with Boundary-Aware Information Bottleneck](https://mlsanity.com/paper/2604.05526/) (2026)
- [Sparsity-Aware Voxel Attention and Foreground Modulation for 3D Semantic Scene Completion](https://mlsanity.com/paper/2604.05780/) (2026)
- [SPARK: Susceptibility-Guided Profiling and Steering of Latent Reasoning States in Large Language Models](https://mlsanity.com/paper/2607.10296/) (2026)
- [Buried in Textual Debt: Context Pruning with Visual Evidence Preservation for MLLM Agents](https://mlsanity.com/paper/2608.22963/) (2026)
- [Motion-Saliency Complementary Masked Modeling for Point Cloud Video Understanding](https://mlsanity.com/paper/2608.30279/) (2026)
- [DRT: Dense Reasoning Trace for Efficient and Grounded Multimodal Reasoning](https://mlsanity.com/paper/2609.21675/) (2026)

## Abstract (arXiv)

While Chain-of-Thought (CoT) reasoning significantly enhances the performance of Multimodal Large Language Models (MLLMs), its autoregressive nature incurs prohibitive latency constraints. Current efforts to mitigate this via token compression often fail by blindly applying text-centric metrics to multimodal contexts. We identify a critical failure mode termed Visual Amnesia, where linguistically redundant tokens are erroneously pruned, leading to hallucinations. To address this, we introduce V-Skip that reformulates token pruning as a Visual-Anchored Information Bottleneck (VA-IB) optimization problem. V-Skip employs a dual-path gating mechanism that weighs token importance through both linguistic surprisal and cross-modal attention flow, effectively rescuing visually salient anchors. Extensive experiments on Qwen2-VL and Llama-3.2 families demonstrate that V-Skip achieves a 2.9× speedup with negligible accuracy loss. Specifically, it preserves fine-grained visual details, outperforming other baselines over 30% on the DocVQA.