Junqing Du, Fernando Ropero, Erkin Turkoz +2cs.CV cs.AI cs.RO
3D spatial reasoning underpins understanding and acting in the physical world, yet it remains unreliable in current multimodal large language models (MLLMs). These models falter at precise geometric measurement, at transforming between egocentric and allocentric viewpoints, and at grounding fine-grained appearance. The most common remedies fine-tune the model on large-scale curated spatial-reasoning datasets or attach dedicated encoders for 3D geometry, which typically couples the solution to costly supervision and a specific backbone. We instead introduce GraFT, a training-free framework that supplies the missing 3D structure through a compact, easily maintained 3D scene graph (3DSG). From this 3DSG, GraFT provides three spatial reasoning capabilities: (1) deterministic geometry through symbolic tools, (2) allocentric layout through a bird's-eye-view (BEV) rendering, and (3) visual-attribute grounding through task-relevant egocentric frames. On ScanQA, GraFT improves every metric over the same-backbone baseline, raising CIDEr by 27%. On VSI-Bench, GraFT improves frozen MLLMs by up to 65%, surpassing every proprietary and general-purpose open-source baseline, and several prominent fine-tuned spatial models.
Composed Image Retrieval (CIR) is an emerging paradigm in content-based image retrieval that enables users to formulate compositional queries by combining a reference image with an auxiliary modality, usually text-based. This approach supports fine-grained search where the target image shares structural elements with the user-provided image while incorporating the modifications specified by the auxiliary text. Conventional CIR methods rely on multimodal fusion to combine visual and textual features into a joint query embedding, which requires training modules that align composed queries with the targets. In this work, we propose PeFuse (for pseudo-fusion), a training-free framework that leverages pretrained Diffusion Models and Multimodal Large Language Models to bridge modalities via generative conversion. We introduce two novel strategies: uni-directional and bi-directional conversion, which convert CIR into four single-modality retrieval problems. These methods reformulate CIR as either intra-modal or cross-modal single-query retrieval tasks, bypassing the need for dedicated task-specific training. Extensive experiments on standard benchmarks demonstrate that converting CIR into text-to-image retrieval tasks is more effective than alternative conversion strategies, achieving competitive or superior performance compared with state-of-the-art methods, while maintaining high flexibility thanks to replaceable components of the conversion pipeline. These results highlight the effectiveness of the pseudo-fusion paradigm for zero-shot CIR. Our code is publicly available at: https://github.com/StevenXuf/PeFuse4CIR.
Vision-language models can be personalized in a training-free manner by directly providing user profiles, preferences, or visual references at inference time, without updating model parameters. However, direct personalized prompting does not guarantee that the model will reliably exploit such evidence. The predictive distribution under the positive user profile often mixes two sources: personalized signals genuinely supported by the current profile, and the model's generic visual or linguistic priors. As a result, from the positive-profile response alone, it is difficult to determine whether a high-confidence answer is supported by the user profile or merely reflects the model's default preference. To address this problem, we propose a training-free calibrated residual decoding framework. Given the same image and question, we construct three evidence conditions: a positive profile , a counterfactual profile , and an empty profile . Our method keeps the prediction under as the anchored base, and explicitly estimates the marginal contribution of personalization from score differences across the three conditions. We further introduce normalized-entropy-based uncertainty calibration, allowing the strength of personalized enhancement to adapt to the reliability of the residual signal. Experiments on MMPB, YoLLaVA, and MyVLM show that the proposed method improves personalized multimodal understanding without fine-tuning, with consistent gains on identity-sensitive visual personalization tasks. Additional analysis shows that entropy calibration stabilizes residual decoding when the contrastive personalization signal is uncertain.
Micro-actions are subtle, short, low-amplitude body movements, such as a fidgeting hand or a slight head tilt, that humans perform with little conscious intent yet that reliably leak emotional and psychological state. Understanding them goes beyond assigning a label: a model must also describe which body parts move and reason, faithfully, about why a clip warrants a particular fine-grained category. We present the training-free, prompt-only system that won first place in the fine-grained understanding track (MA-Bench) of the MAC~2026 Micro-Action Challenge, where both fine-tuning and ground-truth supervision are disallowed. Built entirely upon frozen multimodal large language models (MLLMs), the system dynamically routes each of the eight sub-tasks to the MLLM empirically best suited for that task: a discriminative MLLM for closed-ended recognition tasks and a generative MLLM for open-ended description and reasoning tasks. This architecture achieves a statistically significant performance advantage on open-ended tasks, attaining an average score of 2.68 (on a five-point scale) compared to 1.44 for the second-best approach.
LLaVA-style Vision-Language Models (VLMs) pass visual tokens from a fixed late layer of the vision backbone, typically the penultimate one, to the language model. We first show that this hidden convention is fragile: across 2 VLMs and 7 image and video benchmarks, the default layer is sub-optimal in 13 of 14 model-task pairs, and the best layer shifts with both task and visual backbone. Finding that layer by exhaustive layer-wise inference is prohibitively expensive, and no better fixed default exists. We therefore ask whether layer usefulness can instead be predicted from representation geometry. We study matrix-based entropy, introduced for unimodal layer analysis, which we compute over sample-level visual embeddings as Visual Dataset Entropy (VDE); and Gromov-Wasserstein (GW) distance, introduced for encoder-level VLM model selection, which we repurpose as a layer-wise visual--language alignment signal. Transferring these to LLaVA-based models is not obvious a priori: the vision tower is frozen while the multimodal projector is trained, so we profile both sides of the projector. We find that VDE transfers, and GW does not. Computed from 100 unlabeled task samples without downstream inference, pre-projector VDE tracks layer-wise accuracy and its top-ranked layers cover the oracle best layer on every task for the SigLIP-based LLaVA-Video, while giving region-level guidance for the CLIP-based Video-LLaVA. Post-projector profiles show that the projector reshapes visual geometry but does not erase the performance-relevant trend, leaving $\mathrm{VDE}_{\mathrm{pre}}$ the stronger signal. GW instead flattens after projection and is best read as an alignment diagnostic rather than a selector. VDE thus offers an interpretable, training-free policy that narrows the visual-layer search to a handful of candidates for limited downstream verification.
Multimodal large language models often generate reasoning chains containing subtle errors that lead to incorrect answers. Current verification approaches have notable limitations. Existing approaches either require expensive labelled supervision with inconsistent cross-task performance or aggregate scores from multiple sources by simple aggregations, missing a key insight: when these scores disagree, that disagreement itself carries important information about whether a reasoning step is truly valid or not. We formalise this as a coupled scoring problem among disparate, frozen verifiers, interpretable as a coordination game with a unique closed-form equilibrium where agreement signals valid steps while disagreement reveals instability. Towards this end, we propose a training-free domain-agnostic step-wise verification approach we call VERDICT: VERification via Disagreement-Informed Coupled Thresholding. To our knowledge, VERDICT is the first training-free verifier that makes the structure of cross-modal disagreement explicit and actionable. It computes consensus scores through a closed-form solution, enabling both disagreement-aware filtering and stability-conscious ranking of reasoning steps. Evaluated across six benchmarks, \method consistently improves over the base model by up to +5.95%, and performs competitively with domain-specific critics that demand extensive supervision, demonstrating that cross-modal agreement provides robust verification signals without task-specific adaptation and Training-Free Verification
Ali Cheraghian, Hamidreza Dastmalchi, Hamed Barzamini +4cs.CV
Recent advances in large vision-language models (LVLMs) have enabled powerful multimodal reasoning by integrating visual encoders with large language models (LLMs). However, their reliability is frequently undermined by hallucinations, where generated text inaccurately describes the visual input. Although fine-tuning can mitigate this problem, it is computationally expensive and requires large, curated datasets, making training-free alternatives attractive. Among these, model editing is more promising than decoding-based approaches: decoding methods adapt outputs per input but introduce computational overhead and instability, whereas model editing modifies internal representations offline, providing a more efficient and stable solution. However, existing model-editing techniques typically rely on a single global subspace to correct hallucinations, treating all test samples identically and failing to capture diverse hallucination modes across inputs. To address this limitation, we propose a training-free hallucination mitigation framework for dynamic, per-instance suppression at test time. Our method first constructs a set of Disentangled Hallucination Subspaces, each isolating a distinct hallucination mode. During inference, the model adaptively calculates weights reflecting each input's relationship to these subspaces, guiding a dynamically combined projection that selectively suppresses the most probable hallucination directions while preserving image-grounded semantics. Extensive experiments across multiple vision-language benchmarks and LVLM families demonstrate consistent improvements, highlighting the robustness, generalizability, and efficiency of our approach.
Multimodal LLMs that recognise events reliably still fail to say when they happen. Prompted for timestamps, strong VLMs reach as little as $3.8\%$ R@0.5 on Charades-STA, and $77$ to $80\%$ of their wrong predictions carry low output entropy: the models are confidently wrong, and entropy-based error detection stays below a random classifier. We show that this failure lives in the task interface, not in perception. Holding the weights fixed, replacing timestamp regression with a coarse-to-fine scan of binary questions, whose first-token probabilities are consumed only as a ranking, raises R@0.5 by $28$ to $50$ points across four frozen backbones. The residual failures decompose into two measurable axes: a perception axis that moves with the backbone, and a geometry axis that is analytically predictable from the ratio of the output-window and event widths. FV-Action, the training-free method built on this analysis, reaches $56.8\%$ R@0.5 on Charades-STA, above the same backbone's native grounding pipeline and the strongest training-free result on this benchmark; it surpasses every TVG-trained model evaluated zero-shot on TACoS, and improves over direct prediction on ActivityNet Captions and QVHighlights, with no temporal supervision at any stage.
Vision-language models (VLMs) excel at open-ended captioning and visual QA but often describe objects, attributes, or relations absent from the image, a phenomenon known as object hallucination. We propose a {training-free, post-hoc representation editing technique} that operates in the representation space of the language backbone. The method performs a lightweight, one-time offline calibration on a modest paired dataset to estimate the required covariance structures, using only forward passes and empirical second-order statistics with no gradient updates or fine-tuning, after which the correction is absorbed directly into the model's existing weights. By modeling hidden states as a superposition of truthful and hallucination-associated components, we derive a Wiener-type estimator whose optimal gains are given in closed form from the covariances of paired truthful and hallucinated representations. An eigendecomposition yields mode-wise attenuation that respects a stability criterion, i.e., the filter responds continuously to estimation noise. The correction is applied once to the feed-forward output projections of selected deeper layers, at inference time, the model runs unchanged and at the same speed. Experiments on LLaVA-1.5, MiniGPT-4, Gemma3, and mPLUG-Owl2 demonstrate consistent reductions in object hallucination on CHAIR, POPE, and MME while maintaining caption fluency and overall response quality. We further demonstrate the generality of our approach on the TempCompass video understanding benchmark and on discrete diffusion language models for grounded dialogue, showing that representation filtering reduces hallucinations even in temporal video reasoning and multi-step, sequence-wide denoising settings.
Jaemo Jeong, Junho Yoon, Hyunju Kim +1cs.CV cs.MM cs.SD
Audio-visual event perception (AVEP) determines which events occur in a video, when they occur, and whether they are audible, visible, or both. Training-free methods query new event vocabularies by matching frozen audio and visual features with text-encoded event names. However, related labels share evidence. An incorrect label can then score at least as high as a correct one. We call this a false co-activation (FCA). No scalar cutoff can reject the incorrect label while keeping every correct one. Class-specific thresholds may prevent that label from becoming a final prediction, but the FCA remains in the underlying score vector. We introduce SCoPE, a training-free framework in which all queried labels compete for shared evidence and each modality guides event selection in the other. We derive an exact condition for when this competition removes an FCA in a two-label fit. With identical frozen CLIP+CLAP backbones on LLP, SCoPE improves Type@seg by 7.45 points and Event@seg by 5.04 points compared with the reported AV$^2$A values. The same fixed configuration transfers unchanged to OV-AVEBench and VGGSound-AVEL100k.
Although Multimodal Large Language Models (MLLMs) have made substantial progress, their spatial reasoning may still produce intermediate judgments inconsistent with the input image, allowing errors to propagate through the reasoning chain and affect the final answer. Existing methods mainly improve spatial reasoning through training or additional spatial information, without considering whether the reasoning process itself is faithful to the model input. Our study shows that unfaithful reasoning chains significantly reduce final-answer accuracy. To address this issue, we propose a modular and training-free framework for spatial reasoning verification and correction. The framework constructs a Spatial Evidence Graph (SEG), which associates atomic spatial evidence extracted from Chain-of-Thought reasoning with visual entities, spatial relations, source steps, and visual evidence. Spatial Evidence Reliability Assessment (SERA) evaluates the reliability of visual evidence based on object existence, localization, and geometric measurements. The framework then identifies the earliest spatial evidence unit contradicted by reliable visual evidence and guides the original MLLM to revise the subsequent reasoning and final answer. Across 15 model-dataset settings, our method achieves an average accuracy of 68.94%, outperforming the compared baselines by 8.55 percentage points on average. Our code will be open-sourced.
Recent advances in large language models (LLMs) and vision-language models (VLMs) have enabled new possibilities for 3D question answering (3D-QA), a key capability for embodied AI and robotic perception. However, most existing methods rely on 3D-specific training or fine-tuning with costly annotations, limiting their scalability and real-world applicability. We present \textbf{ViewMind3D}, a fully training-free and modular framework for 3D spatial reasoning over multi-view observations of a scene without requiring complete 3D reconstruction. The framework decomposes the 3D-QA task into four interpretable components: (1) question-driven multi-view selection, (2) guided visual grounding with language-conditioned object cues, (3) spatial context encoding via a bird's-eye-view (BEV) viewpoint indicator, and (4) structured answer generation through role-based reasoning. This design enables structured, robust, and interpretable reasoning without requiring model tuning. Experimental results on ScanQA and SQA3D show that ViewMind3D achieves competitive performance compared to prior training-free and fine-tuned 3D-LLMs. In particular, our method improves performance on spatially grounded question types, such as ``What'' questions in SQA3D, while maintaining strong overall accuracy (50.8\%) and achieving 73.4 CIDEr on ScanQA. These results demonstrate that effective 3D reasoning can be achieved through modular orchestration of general-purpose LLMs and VLMs for robotic perception in real-world environments.
High-resolution visual question answering (HR-VQA) is often treated as a problem of insufficient evidence acquisition, where failing multimodal large language models must inspect images again through cropping, re-encoding, or multi-round search. We show that this view is incomplete: in many cases, fine-grained evidence has already survived visual encoding and become identifiable and influential within an intermediate-layer routing window, but is later diluted before answer generation. We propose Thinking-Once, a \textbf{training-free, single-visual-pass} evidence-routing method that reconstructs question-conditioned attention at this window, preserves core entity tokens and compact background context, and routes this evidence to later layers without extra visual encoding. Across five base models, Thinking-Once consistently improves or matches the corresponding base setting, increasing the average scores on V$^*$Bench, HRBench-4K, and HRBench-8K by \textit{+3.1}, \textit{+3.0}, and \textit{+2.7} points while reducing the average peak memory by about 4,GB. On Qwen2.5-VL-7B, it improves the three benchmarks by \textit{+9.9}, \textit{+4.6}, and \textit{+5.5} points, raising the cross-benchmark mean from 72.5 to 79.1. With the ZwZ-8B base model, Thinking-Once reaches a mean score of 82.7. Against 11 open-source HR-VQA baselines, it obtains the best or tied-best score on all three benchmark averages and the best overall mean; for example, compared with DeepScan, it reduces V$^*$Bench inference time by \textbf{97.2\%} while improving the cross-benchmark mean from 77.8 to 79.1. These results show that HR-VQA can be improved by routing already encoded evidence rather than repeatedly acquiring new visual inputs. Code is available in the appendix.
Understanding long videos with multimodal large language models (MLLMs) requires selecting a compact set of frames from thousands of candidates, yet identifying the right frames seemingly requires understanding the video first. We resolve this circular dependency with a simple observation: cross-modal attention at validation-selected extraction layers in MLLMs already provides query-relevant frame evidence without requiring autoregressive generation. We exploit this property to build DAFS (Dynamic Attention-based Budget-aware Frame Selection), a training-free frame selector. A lightweight MLLM selector, even with only 2B parameters, can extract frame-level evidence by converting selected-layer attention into relevance scores through query-conditioned aggregation. This enables cross-frame comparison without autoregressive decoding. To handle the selector's own context constraint, we formulate the joint allocation of candidate pool size and per-frame token budget as a discrete optimization problem solved by dynamic programming. Under a 32-frame budget, our selector improves over uniform sampling by up to 6.4 points on Video-MME and outperforms prior training-based selectors under matched frame budgets, while generalizing across selector and answerer backbones, and across tasks, without retraining.
Rakshanda Hassan Abhinandan, John Galeotti, Deva Ramanan +1cs.CV cs.AI cs.LG eess.IV
Where should the question go in a vision-language model (VLM) prompt: before the image or after it? Intuition says before: knowing what is asked should tell the model where to look. Yet across visual question answering benchmarks, question-first prompting consistently underperforms the image-first ordering recommended for frontier VLMs, a phenomenon we term the question-first paradox. We trace this paradox to a conflict between two stages of VLM computation. Logit-lens and attention probes show that question-first prompting steers perception, shifting image patch representations toward question-relevant concepts. But downstream, stranded behind hundreds of image tokens, the question is barely attended by the answer token, which instead commits to image-driven, often wrong answers. Causal attention knockout confirms that the answer reads the question only when it follows the image. This diagnosis yields a training-free fix: question echoing, restating the question on both sides of the image so one copy steers perception while the other is available at answer time. A similar division of labor appears in a fifty-year-old finding on human 'adjunct questions', where repeating a question before and after a passage improves comprehension. Echoing the image as well brings further gains by restoring the whole-image view otherwise lost by a causal decoder. The paradox holds across five open VLMs, costing up to 17.5 group-accuracy points. Echoed prompts recover most of the gap and, on NaturalBench and Winoground, surpass the best single-pass ordering by up to 19 group-accuracy points on Winoground, with no training, fine-tuning, or architecture change. The paradox reveals a tension between steering what a model sees and preserving access to what it was asked; echoing resolves this through prompt design. Project Page: https://rakshanda-cmu.github.io/ask-twice-look-twice/
In this paper, we tackle the Aerial Vision-and-Dialog Navigation (AVDN) task in the training-free setting for resource-efficient high-altitude UAV navigation.Naively applying MLLMs leads to unreliable navigation due to weak directional grounding and the lack of explicit spatial memory.To address these issues, we propose PSC-AVDN, a training-free framework that tightly couples a three-stage Parsing-Search-Confirmation reasoning pipeline with a Structured Spatial Memory (SSM).The parsing stage uses an LLM to convert ambiguous dialogue instructions into stable geometric directional and destination cues.A Search Chain-of-Thought (S-CoT) then performs stepwise target exploration under high-altitude observations, and a Confirmation Chain-of-Thought (C-CoT) conducts fine-grained verification around candidate regions to resolve visual ambiguity.Meanwhile, SSM integrates three complementary sources of spatial cues, including multi-scale visual observation, spatial visual memory, and structured geometric memory to provide global spatial context and long-horizon consistency.Extensive experiments on ANDH and ANDH-Full show that PSC-AVDN establishes new state-of-the-art performance in the training-free setting, matching or surpassing several finetuned methods.Code will be publicly available at: https://github.com/QY6616/PSC-AVDN
Large Vision-Language Models (LVLMs) have achieved remarkable progress in visual understanding tasks such as image captioning and visual question answering. However, they remain susceptible to hallucinations, generating content that is inconsistent with the actual visual input. Existing methods primarily intervene at the decoding stage, while overlooking a critical source of hallucinations: irrelevant or noisy visual tokens that mislead the decoding process. To address this issue, we propose SeeMe, a training-free framework that introduces the concept of feature engineering from traditional machine learning into LVLMs. SeeMe restructures visual tokens through a three-stage token engineering process to suppress hallucination sources while preserving informative visual evidence. Experiments on MME, POPE, and AMBER benchmarks across four LVLMs demonstrate that SeeMe consistently reduces hallucinations and improves output consistency, providing a novel perspective for mitigating hallucinations in LVLMs.
As grounded QA systems are increasingly deployed in AI assistants, accurately attributing generated answers to evidence is critical for user trust and model safety. While unimodal attributions have been explored in depth, the multimodal setting remains relatively under-researched. As a result, we introduce MultAttnAttrib, a training-free attribution-generation method that leverages a model's prefill pass, selected attention heads, and calibrated thresholds to locate source evidence within a document. To establish baseline results for the method, we introduce MultAttrEval, a complementary benchmark dataset annotated with fine-grained, ground-truth attributions for answer components grounded in multimodal source documents. To our knowledge, this is the first evaluation dataset designed specifically for multimodal attribution in long-form documents. Experimental results show that MultAttnAttrib consistently outperforms a variety of attribution-generation methods, including several strong prompting-based approaches and matches the latest frontier models such as GPT 5.4. Our method not only substantially improves attribution accuracy for both unimodal and multimodal attribution types, but also produces attributions at up to one-seventh of the direct inference latency compared to prompting on the same base model.
High-resolution (HR) image perception remains a key challenge in multimodal large language models (MLLMs), as fine-grained details are often lost when the image is processed as a whole. Existing methods either require training to teach models where to look or heuristically divide the image into fixed regions, both of which struggle to generalize in complex HR scenes. In this work, we propose Hierarchical Entity Exploration (HEE), a training-free and model-agnostic framework that transforms static image understanding into dynamic, query-guided entity exploration. HEE first evaluates each region using a dual scoring mechanism to determine whether it already contains sufficient evidence to answer the question. If not, it applies object detection within the most promising region to extract fine-grained entities, clusters them into coherent subregions, and organizes them into a multi-level semantic hierarchy for deeper exploration. When deeper regions still fail to yield confident answers, a confidence-guided backtracking mechanism revisits alternative paths to ensure adaptive perception. Extensive results show that HEE outperforms training-free methods like ZoomEye and RAP in both accuracy and efficiency on two complex HR benchmarks (Visual Probe and HR-Bench), across different MLLMs such as Qwen2.5-VL and LLaVA-OneVision. Moreover, HEE demonstrates generalization on the MME-RealWorld benchmark.
Remote sensing visual grounding (RSVG) aims to locate specific objects in high-resolution RS imagery using free-form natural language descriptions. While recent advances in multimodal large language models (MLLMs) show great potential for such open-vocabulary RSVG, their training-free adaptation is hindered by the modality gap between abstract linguistic semantics and fine-grained visual cues. In cluttered RS scenes, this gap inevitably causes severe localization drift. To bridge this gap, we propose Exemplar-driven Calibrated Refinement (ExACT), a novel training-free framework driven by a one-shot visual prompting mechanism to explicitly provide discriminative structural guidance for precise pixel-level localization. Specifically, we propose a Vision Exemplar-based Calibrator (VEC) that extracts fine-grained visual correspondences from the given exemplar to rectify the rough cross-modal priors from frozen MLLMs, effectively suppressing background artifacts and accurately outlining target boundaries. Subsequently, a Structure-Aware Refiner (SAR) employs an iterative merge-and-select clustering strategy to consolidate the calibrated priors into high-quality positive and negative geometric prompts. These prompts then guide the Segment Anything Model (SAM) to achieve precise pixel-level predictions. Extensive experiments confirm the superiority of ExACT over existing training-free and weakly-supervised methods.
Multimodal Large Language Models (MLLMs) have demonstrated impressive vision-language understanding, yet still struggle with fine-grained perception in high-resolution images. While existing training-free methods typically rely on attention-based localization or coarse-to-fine search, they are often misled by distractors and fail to locate multiple targets. Our investigation attributes these failures to Contextual Dominance, where salient distractors overwhelm target attention and cause inaccurate localization, and Semantic Bias, where global semantics cause the model to fixate on the most salient concept, resulting in incomplete localization in multi-object scenarios. Built on these insights, we propose ActiveScope, a training-free framework that enhances MLLMs by actively seeking and correcting perception. ActiveScope features two modules. The Semantic Anchor Localization (SAL) utilizes fine-grained semantic anchors to independently localize key targets, thereby mitigating semantic bias. The Interference-Suppressed Refinement (ISR) refines localization by suppressing attention on salient distractions to overcome contextual dominance. Extensive experiments on high-resolution image understanding benchmarks demonstrate that ActiveScope outperforms existing training-free methods (e.g., 96.34 percent accuracy on $V^{*}$ Bench), validating the superiority of the active search and self-correction paradigm. Our code is available at https://github.com/jasmine-ww/ActiveScope.
Knowledge-Based Visual Question Answering (KB-VQA) requires grounding visual queries to external knowledge beyond directly observable content in images. While recent multi modal large language models (MLLMs) show strong perceptual abilities, they struggle on KB-VQA tasks requiring groundings from both fine-grained entity and evidence levels. Most existing multi-modal retrieval augmented generation (MM-RAG) methods tightly couple entity discrimination and section-level evidence ranking into a single re-ranking stage, leading to high cost and limited generalization. In this work, we revisit existing MM-RAG solutions from a workflow perspective and argue both entity-level and fact-level groundings are key bottlenecks. We observe that although MLLMs often fail under open-ended entity naming, they can better identify the correct entity when selecting from a small set of candidate names. Based on this insight, we propose a simple and training-free identify-before-answer IBA framework that decouples entity identification from section-level re-ranking. Our approach prompts an MLLM to select high-confidence entities using only candidate names, followed by an off-the-shelf textual re-ranker for evidence selection. Experiments on Encyclopedic-VQA and InfoSeek show that our method consistently outperforms fine-tuned multi-modal re-ranking baselines while reducing training and inference complexity. Additional analyses reveal that the improvements arise not only from better entity identification, but also from selecting more informative evidence once correct entity is fixed. Our implementation is made public to ease reproducibility.
Vision-language models (VLMs) enable visual recognition from semantic class descriptions, which makes them attractive when target annotations are scarce or unavailable. Most deployment pipelines, however, first choose a single VLM and then adapt that model to the unlabeled target set. This single-backbone paradigm hides a critical assumption: the selected VLM is already compatible with the target domain. In realistic cross-domain deployment, several general-purpose and domain-specialized VLMs may be plausible, yet no instance-level target labels are available to identify the reliable ones. Deployment therefore requires a coupled solution for model selection, target adaptation, and prediction integration. We revisit this problem from a system-level multi-VLM perspective. Our central observation is that the three decisions above depend on the same latent object: a trustworthy sample-class structure in the target set. Different VLMs may encode different transfer biases and produce conflicting predictions, but their outputs can still provide complementary evidence for estimating this structure. We propose One Stone, Three Birds, a training-free framework based on self-adaptive optimal transport. Given a pool of frozen candidate VLMs, OSTB estimates a consensus sample-to-class transport plan without updating VLM parameters. The learned transport structure is then reused for all deployment objectives: model selection is performed by ranking the combined semantic and visual reliability induced by the consensus plan; target adaptation is obtained by fitting transport-conditioned visual classifiers; and ensembling is implemented through reliability-aware probabilistic integration. Extensive experiments on natural-image, remote-sensing, and medical-pathology benchmarks show that OSTB improves model ranking, adaptation stability, and ensemble robustness under heterogeneous candidate pools.
Recent advances in large vision-language models have expanded video retrieval from simple text-based search to more flexible scenarios, where users may specify the desired result through both visual examples and textual instructions. In the CVPR 2026 Reason-Aware Composed Video Retrieval Challenge, the system is required to retrieve a target video according to a reference video and a modification instruction. To address this task, we develop Visual Representation-Guided Video-LLM Reasoning for Training-Free Composed Video Retrieval. Our framework first uses frozen DINOv3 models to obtain a compact set of visually relevant candidates, and then applies large vision-language models to evaluate whether each candidate satisfies the modification instruction. A final reasoning-based refinement is further performed on the top candidates to improve the first-ranked prediction. Without training, our system achieves 48.78 Recall@1 and 51.48 Recall@5 on the test set. Future work may further improve retrieval accuracy through stronger video-LLMs and detailed integration between visual representations and language reasoning.
Recent advances in Multimodal Large Language Models (MLLMs) have significantly advanced video understanding tasks, yet challenges remain in efficiently compressing visual tokens while preserving spatiotemporal interactions. Existing methods, such as LLaVA family, utilize simplistic pooling or interpolation techniques that overlook the intricate dynamics of visual tokens. To bridge this gap, we propose ST-GridPool, a novel training-free visual token enhancement method designed specifically for Video LLMs. Our approach integrates Pyramid Temporal Gridding (PTG), which captures multi-grained spatiotemporal interactions through hierarchical temporal gridding, and Norm-based Spatial Pooling (NSP), which preserves high-information visual regions by leveraging the correlation between token norms and semantic richness. Extensive experiments on various benchmarks demonstrate that ST-GridPool consistently enhances performance of Video LLMs without requiring costly retraining. Our method offers an efficient and plug-and-play solution for improving visual token representations. Our code is available in https://github.com/bingjunluo/ST-GridPool.
Training-free zero-shot composed image retrieval models are recently gaining increasing research interest due to their generalizability and flexibility in unseen multimodal retrieval. Recent LLM-based advances focus on generating the expected target caption by exploring the compositional ability behind the LLMs. Although efficient, we find that 1) the generated captions tend to introduce unexpected features from the reference image due to the semantic gap between the input image and text modification, where the image contains much more details than the text; 2) the point-to-point alignment during the retrieval stage fails to capture diverse compositions. To address these challenges, we introduce a novel Semantic Transition and Transportation in collaboration framework for training-free zero-shot CIR tasks. Specifically, given the composed caption inferred by an LLM, we aim to refine it through a transition vector in the embedding space and make it closer to the target image. Combining LLMs with user instruction, the refined caption concentrates more on the core modification intent and thus filters out unnecessary noise. Moreover, to explore diverse alignment during the retrieval stage, we model the caption and image as discrete distributions and reformulate the retrieval task as a set-to-set alignment task. Finally, a bidirectional transportation distance is developed to consider fine-grained alignments across modalities and calculate the retrieval score. Extensive experiments demonstrate that our method can be general, effective, and beneficial for many CIR tasks.
Large Vision-Language Models (LVLMs) have shown remarkable performance on a wide range of vision-language tasks. Despite this progress, they are still prone to hallucination, generating responses that are inconsistent with visual content. In this work, we find that LVLMs tend to hallucinate when they pay insufficient attention to the correct visual evidence and gradually forget it during the generation process. We empirically find that although LVLMs overall attend insufficiently to visual evidence, they exhibit sensitivity to the correct visual evidence in specific layers, with notable inter-layer discrepancy. Motivated by this observation, we propose a novel hallucination mitigation method that enhances visual evidence based on Inter-Layer Visual Attention Discrepancy (ILVAD). Specifically, we obtain the attention weights from early generated tokens to visual tokens across layers and identify the tokens that are repeatedly activated as visual evidence, forming a saliency map. We then enhance attention to visual evidence during generation through the saliency map to reduce visual forgetting. In addition, we leverage the saliency map to obtain attention scores of generated text to visual evidence, in order to select and emphasize text tokens that are strongly grounded in visual evidence. Our method is training-free and plug-and-play. Multiple benchmark evaluations conducted on five recently released models show that our method can consistently mitigate hallucinations in different LVLMs over various architectures. Code is available at https://github.com/ytx-ML/ILVAD.
Although Large Vision-Language Models (LVLMs) have demonstrated remarkable performance on downstream tasks, they frequently produce contents that deviate from visual information, leading to object hallucination. To tackle this, recent works mostly depend on expensive manual annotations and training cost, or decoding strategies which significantly increase inference time. In this work, we observe that LVLMs' attention to visual information is significantly enhanced when answering caption queries compared to non-caption queries. Inspired by this phenomenon, we propose Caption-guided Visual Attention Steering (CAST), a training-free, plug-and-play hallucination mitigation method that leverages the attention activation pattern corresponding to caption queries to enhance LVLMs' visual perception capability. Specifically, we use probing techniques to identify attention heads that are highly sensitive to caption queries and estimate optimized steering directions for their outputs. This steering strengthens LVLM's fine-grained visual perception capabilities, thereby effectively mitigating object hallucination. CAST reduced object hallucination by an average of 6.03% across five widely used LVLMs and five benchmarks including both discriminative and generative tasks, demonstrating state-of-the-art performance while adding little inference cost and preserving other foundational capabilities.