Challenge: Multimodal large language models have strong performance on visual question answering benchmarks . however, their inference efficiency is severely constrained by the rapidly growing context .
Approach: They propose a modality-decoupled compression method that enables efficient multimodal inference . they propose to evict visual tokens whenever visual grounding is unnecessary .
Outcome: The proposed method reduces the average context length by up to 57% while maintaining comparable performance to the standard MLLM baseline.

Similar Papers

Reasoning in the Dark: Interleaved Vision-Text Reasoning in Latent Space (2026.findings-acl)

Copied to clipboard

Challenge: Existing multimodal reasoning methods depend on explicit reasoning steps that require labor-intensive vision-text annotations and inherently introduce significant inference latency.
Approach: They propose a method that integrates visual and visual information into the reasoning process to improve the performance of multimodal LLMs.
Outcome: The proposed method achieves an average performance increase of 5.45% while achieving a speed increase of over 5 times compared to existing methods.
Multimodal Causal Reasoning Benchmark: Challenging Multimodal Large Language Models to Discern Causal Links Across Modalities (2025.findings-acl)

Copied to clipboard

Challenge: Existing MLLMs lack robustness in multimodal causal reasoning compared to their performance in textual settings.
Approach: They propose a novel multimodal chain-of-thought (CoT) reasoning benchmark that leverages siamese images and text pairs to challenge MLLMs.
Outcome: The proposed benchmark leverages siamese images and text pairs to challenge MLLMs.
Learning Flexible Large Multimodal Models with Arbitrary Modality Combinations (2026.findings-acl)

Copied to clipboard

Challenge: Multimodal Large Language Models (MLLMs) have potential for cross-modal understanding . but extending MLLM to handle diverse modalities introduces two challenges .
Approach: They propose a dual-stage compression mechanism to reduce the number of modality tokens per modality and condense it into a single, compact token sequence.
Outcome: Experiments show that Flex-M3 outperforms its counterpart trained on only full-modality data.
Reducing Peak Memory Usage for Modern Multimodal Large Language Model Pipelines (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods to reduce memory usage of multimodal large language models rely on storing large numbers of vision tokens in the key–value cache . however, such compression is typically only applied after all inputs are processed, resulting in high peak memory usage during the prefill stage.
Approach: They propose a sequential input-compression mechanism that enforces a fixed memory budget by performing structure-aware key–value cache compression during the prefill stage.
Outcome: The proposed approach reduces peak memory usage while maintaining generative performance with only minimal degradation, enabling more practical and memory-efficient multimodal inference.
VisiPruner: Decoding Discontinuous Cross-Modal Dynamics for Efficient Multimodal LLMs (2025.emnlp-main)

Copied to clipboard

Challenge: Multimodal Large Language Models (MLLMs) suffer from significant computational overhead due to the quadratic growth of attention computations with the number of multimodal tokens.
Approach: They propose a training-free pruning framework that prunes multimodal tokens without a trained pruning method.
Outcome: The proposed pruning framework outperforms existing token pruning methods and generalizes across diverse MLLMs.
Chain-of-Thought Compression Should Not Be Blind: V-Skip for Efficient Multimodal Reasoning via Dual-Path Anchoring (2026.acl-long)

Copied to clipboard

Challenge: Existing efforts to mitigate this via token compression fail due to its autoregressive nature . linguistically redundant tokens are erroneously pruned, leading to hallucinations .
Approach: They propose a method that reformulates token pruning as a Visual-Anchored Information Bottleneck (VA-IB) optimization problem.
Outcome: Experiments on Qwen2-VL and Llama-3.2 families show that the proposed model achieves a speedup with negligible accuracy loss.
Simple-VGC: Enhancing Visual Grounding in Multimodal Reasoning via Adaptive Tool Composition (2026.acl-long)

Copied to clipboard

Challenge: Existing multimodal large language models suffer from systematic failures in basic visual understanding.
Approach: They propose a tool-augmented reasoning framework with three targeted compensation strategies to address these problems.
Outcome: The proposed framework improves visual grounding by re-injecting the original image to mitigate visual forgetting, the authors show . the proposed framework also improves the accuracy of the visual inputs, the researchers show - and the results are promising .
Less is More: A Simple yet Effective Token Reduction Method for Efficient Multi-modal LLMs (2025.coling-main)

Copied to clipboard

Challenge: Recent advances in Multimodal Large Language Models have led to a significant surge in the resource consumption of these models.
Approach: They propose a method to reduce image tokens using visual query data by using CLIP metrics to reduce computational overhead and maintain consistent performance.
Outcome: The proposed method has been extensively tested across 12 datasets and shows a significant reduction in computational overhead while maintaining a consistent level of performance.
Protecting multimodal large language models against misleading visualizations (2026.acl-long)

Copied to clipboard

Challenge: MLLMs are robust to misleading visualizations, i.e., charts that distort the underlying data, leading readers to draw inaccurate conclusions.
Approach: They propose to use table-based QA and redrawing the visualization to improve QA performance on misleading visualizations.
Outcome: The proposed methods improve MLLM question-answering accuracy on misleading visualizations without compromising accuracy on non-misleading ones.
Probing Audio-Visual Reasoning in Multimodal Language Models through the Lens of Audio (2026.acl-long)

Copied to clipboard

Challenge: Recent multimodal large language models lack robust audio-visual integration ability and performance on DeafTest is highly correlated with AV-Odyssey accuracy.
Approach: They propose a benchmarking tool that integrates audio-visual reasoning with audio-video cues to infer solutions.
Outcome: The proposed model performs well on DeafTest, but lacks audio perception in simple audio tasks.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations