PIO-FVLM: Rethinking Training-Free Visual Token Reduction for VLM Acceleration from an Inference-Objective Perspective

Recently, reducing redundant visual tokens in vision-language models (VLMs) to accelerate VLM inference has emerged as a hot topic. However, most existing methods rely on heuristics constructed based on inter-visual-token similarity or cross-modal visual-text similarity, which gives rise to certain limitations in compression performance and practical deployment. In contrast, we propose PIO-FVLM from the perspective of inference objectives, which transforms visual token compression into preserving output result invariance and selects tokens primarily by their importance to this goal. Specially, vision tokens are reordered with the guidance of token-level gradient saliency generated by our designed layer-local proxy loss, a coarse constraint from the current layer to the final result. Then the most valuable vision tokens are selected following the non-maximum suppression (NMS) principle. The proposed PIO-FVLM is training-free and compatible with FlashAttention, friendly to practical application and deployment. It can be deployed independently as an encoder-free method, or combined with encoder compression approaches like VisionZip for use as an encoder-involved method. On LLaVA-Next-7B, PIO-FVLM retains just 11.1% of visual tokens but maintains 97.2% of the original performance, with a 2.67$\times$ prefill speedup, 2.11$\times$ inference speedup, 6.22$\times$ lower FLOPs, and 6.05$\times$ reduced KV Cache overhead. Our code is available at https://github.com/ocy1/PIO-FVLM.

Model-Dowser: Data-Free Importance Probing to Mitigate Catastrophic Forgetting in Multimodal Large Language Models

Fine-tuning Multimodal Large Language Models (MLLMs) on task-specific data is an effective way to improve performance on downstream applications. However, such adaptation often leads to a degradation in generalization on pretrained tasks, a phenomenon known as Catastrophic Forgetting. Existing methods that aim to mitigate this issue either become ineffective when fine-tuning deeper layers of the language decoder or scale poorly with increasing model size. To address these limitations, we propose Model-Dowser, a novel sparse fine-tuning approach for MLLMs. Model-Dowser measures a principled importance score for each model parameter with respect to pretrained generalization (prior to downstream adaptation) by jointly considering weight magnitudes, input activations, and output sensitivities. During fine-tuning, Model-Dowser selectively preserves high-importance parameters and updates the remaining. Comprehensive experiments on two representative MLLMs, LLaVA and NVILA, demonstrate that Model-Dowser effectively mitigates catastrophic forgetting and consistently outperforms prior methods, while remaining resource-efficient and scalable to multi-billion-parameter models.

When LLaVA Meets Objects: Token Composition for Vision-Language-Models

Current autoregressive Vision Language Models (VLMs) usually rely on a large number of visual tokens to represent images, resulting in a need for more compute especially at inference time. To address this problem, we propose Mask-LLaVA, a framework that leverages different levels of visual features to create a compact yet information-rich visual representation for autoregressive VLMs. Namely, we combine mask-based object representations together with global tokens and local patch tokens. While all tokens are used during training, it shows that the resulting model can flexibly drop especially the number of mask-based object-tokens at test time, allowing to adapt the number of tokens during inference without the need to retrain the model and without a significant drop in performance. We evaluate the proposed approach on a suite of standard benchmarks showing results competitive to current token efficient methods and comparable to the original LLaVA baseline using only a fraction of visual tokens. Our analysis demonstrates that combining multi-level features enables efficient learning with fewer tokens while allowing dynamic token selection at test time for good performance.

Fast-Slow Efficient Training for Multimodal Large Language Models via Visual Token Pruning

Multimodal Large Language Models (MLLMs) suffer from severe training inefficiency issue, which is associated with their massive model sizes and visual token numbers. Existing efforts in efficient training focus on reducing model sizes or trainable parameters. Inspired by the success of Visual Token Pruning (VTP) in improving inference efficiency, we are exploring another substantial research direction for efficient training by reducing visual tokens. However, applying VTP at the training stage results in a training-inference mismatch: pruning-trained models perform poorly when inferring on non-pruned full visual token sequences. To close this gap, we propose DualSpeed, a fast-slow framework for efficient training of MLLMs. The fast-mode is the primary mode, which incorporates existing VTP methods as plugins to reduce visual tokens, along with a mode isolator to isolate the model's behaviors. The slow-mode is the auxiliary mode, where the model is trained on full visual sequences to retain training-inference consistency. To boost its training, it further leverages self-distillation to learn from the sufficiently trained fast-mode. Together, DualSpeed can achieve both training efficiency and non-degraded performance. Experiments show DualSpeed accelerates the training of LLaVA-1.5 by 2.1$\times$ and LLaVA-NeXT by 4.0$\times$, retaining over 99% performance. Code: https://github.com/dingkun-zhang/DualSpeed

ConsensusDrop: Fusing Visual and Cross-Modal Saliency for Efficient Vision Language Models

Vision-Language Models (VLMs) are expensive because the LLM processes hundreds of largely redundant visual tokens. Existing token reduction methods typically exploit \textit{either} vision-encoder saliency (broad but query-agnostic) \textit{or} LLM cross-attention (query-aware but sparse and costly). We show that neither signal alone is sufficient: fusing them consistently improves performance compared to unimodal visual token selection (ranking). However, making such fusion practical is non-trivial: cross-modal saliency is usually only available \emph{inside} the LLM (too late for efficient pre-LLM pruning), and the two signals are inherently asymmetric, so naive fusion underutilizes their complementary strengths. We propose \textbf{ConsensusDrop}, a training-free framework that derives a \emph{consensus} ranking by reconciling vision encoder saliency with query-aware cross-attention, retaining the most informative tokens while compressing the remainder via encoder-guided token merging. Across LLaVA-1.5/NeXT, Video-LLaVA, and other open-source VLMs, ConsensusDrop consistently outperforms prior pruning methods under identical token budgets and delivers a stronger accuracy-efficiency Pareto frontier -- preserving near-baseline accuracy even at aggressive token reductions while reducing TTFT and KV cache footprint. Our code will be open-sourced.

SAGE: Accelerating Vision-Language Models via Entropy-Guided Adaptive Speculative Decoding

Speculative decoding has emerged as a promising approach to accelerate inference in vision-language models (VLMs) by enabling parallel verification of multiple draft tokens. However, existing methods rely on static tree structures that remain fixed throughout the decoding process, failing to adapt to the varying prediction difficulty across generation steps. This leads to suboptimal acceptance lengths and limited speedup. In this paper, we propose SAGE, a novel framework that dynamically adjusts the speculation tree structure based on real-time prediction uncertainty. Our key insight is that output entropy serves as a natural confidence indicator with strong temporal correlation across decoding steps. SAGE constructs deeper-narrower trees for high-confidence predictions to maximize speculation depth, and shallower-wider trees for uncertain predictions to diversify exploration. SAGE improves acceptance lengths and achieves faster acceleration compared to static tree baselines. Experiments on multiple benchmarks demonstrate the effectiveness of SAGE: without any loss in output quality, it delivers up to $3.36\times$ decoding speedup for LLaVA-OneVision-72B and $3.18\times$ for Qwen2.5-VL-72B.

Countering the Over-Reliance Trap: Mitigating Object Hallucination for LVLMs via a Self-Validation Framework

Despite progress in Large Vision Language Models (LVLMs), object hallucination remains a critical issue in image captioning task, where models generate descriptions of non-existent objects, compromising their reliability. Previous work attributes this to LVLMs' over-reliance on language priors and attempts to mitigate it through logits calibration. However, they still lack a thorough analysis of the over-reliance. To gain a deeper understanding of over-reliance, we conduct a series of preliminary experiments, indicating that as the generation length increases, LVLMs' over-reliance on language priors leads to inflated probability of hallucinated object tokens, consequently exacerbating object hallucination. To circumvent this issue, we propose Language-Prior-Free Verification to enable LVLMs to faithfully verify the confidence of object existence. Based on this, we propose a novel training-free Self-Validation Framework to counter the over-reliance trap. It first validates objects' existence in sampled candidate captions and further mitigates object hallucination via caption selection or aggregation. Experiment results demonstrate that our framework mitigates object hallucination significantly in image captioning task (e.g., 65.6% improvement on CHAIRI metric with LLaVA-v1.5-7B), surpassing the previous SOTA methods. This result highlights a novel path towards mitigating hallucination by unlocking the inherent potential within LVLMs themselves.

Gated Relational Alignment via Confidence-based Distillation for Efficient VLMs

Vision-Language Models (VLMs) achieve strong multimodal performance but are costly to deploy, and post-training quantization often causes significant accuracy loss. Despite its potential, quantization-aware training for VLMs remains underexplored. We propose GRACE, a framework unifying knowledge distillation and QAT under the Information Bottleneck principle: quantization constrains information capacity while distillation guides what to preserve within this budget. Treating the teacher as a proxy for task-relevant information, we introduce confidence-gated decoupled distillation to filter unreliable supervision, relational centered kernel alignment to transfer visual token structures, and an adaptive controller via Lagrangian relaxation to balance fidelity against capacity constraints. Across extensive benchmarks on LLaVA and Qwen families, our INT4 models consistently outperform FP16 baselines (e.g., LLaVA-1.5-7B: 70.1 vs. 66.8 on SQA; Qwen2-VL-2B: 76.9 vs. 72.6 on MMBench), nearly matching teacher performance. Using real INT4 kernel, we achieve 3$\times$ throughput with 54% memory reduction. This principled framework significantly outperforms existing quantization methods, making GRACE a compelling solution for resource-constrained deployment.

LLaVA-FA: Learning Fourier Approximation for Compressing Large Multimodal Models

Large multimodal models (LMMs) have achieved impressive performance on various vision-language tasks, but their substantial computational and memory costs hinder their practical deployment. Existing compression methods often decouple low-rank decomposition and quantization, leading to compounded reconstruction errors, especially in multimodal architectures with cross-modal redundancy. To address this issue, we propose LLaVA-FA, a novel efficient LMM that performs joint low-rank plus quantization approximation in the frequency domain. By leveraging the de-correlation and conjugate symmetry properties of Fourier transform, LLaVA-FA achieves more compact and accurate weight representations. Furthermore, we introduce PolarQuant, a polar-coordinate quantization method tailored for complex matrices, and an optional diagonal calibration (ODC) scheme that eliminates the need for large-scale calibration data. Extensive experimental results demonstrate that our proposed LLaVA-FA outperforms existing efficient multimodal models across multiple benchmarks while maintaining minimal activated parameters and low computational costs, validating its effectiveness as a powerful solution for compressing LMMs.

IC-EO: Interpretable Code-based assistant for Earth Observation

Despite recent advances in computer vision, Earth Observation (EO) analysis remains difficult to perform for the laymen, requiring expert knowledge and technical capabilities. Furthermore, many systems return black-box predictions that are difficult to audit or reproduce. Leveraging recent advances in tool LLMs, this study proposes a conversational, code-generating agent that transforms natural-language queries into executable, auditable Python workflows. The agent operates over a unified easily extendable API for classification, segmentation, detection (oriented bounding boxes), spectral indices, and geospatial operators. With our proposed framework, it is possible to control the results at three levels: (i) tool-level performance on public EO benchmarks; (ii) at the agent-level to understand the capacity to generate valid, hallucination-free code; and (iii) at the task-level on specific use cases. In this work, we select two use-cases of interest: land-composition mapping and post-wildfire damage assessment. The proposed agent outperforms general-purpose LLM/VLM baselines (GPT-4o, LLaVA), achieving 64.2% vs. 51.7% accuracy on land-composition and 50% vs. 0% on post-wildfire analysis, while producing results that are transparent and easy to interpret. By outputting verifiable code, the approach turns EO analysis into a transparent, reproducible process.