AVA-Bench: Atomic Visual Ability Benchmark for Vision Foundation Models

The rise of vision foundation models (VFMs) calls for systematic evaluation. A common approach pairs VFMs with large language models (LLMs) as general-purpose heads, followed by evaluation on broad Visual Question Answering (VQA) benchmarks. However, this protocol has two key blind spots: (i) the instruction tuning data may not align with VQA test distributions, meaning a wrong prediction can stem from such data mismatch rather than a VFM' visual shortcomings; (ii) VQA benchmarks often require multiple visual abilities, making it hard to tell whether errors stem from lacking all required abilities or just a single critical one. To address these gaps, we introduce AVA-Bench, the first benchmark that explicitly disentangles 14 Atomic Visual Abilities (AVAs) -- foundational skills like localization, depth estimation, and spatial understanding that collectively support complex visual reasoning tasks. By decoupling AVAs and matching training and test distributions within each, AVA-Bench pinpoints exactly where a VFM excels or falters. Applying AVA-Bench to leading VFMs thus reveals distinctive "ability fingerprints," turning VFM selection from educated guesswork into principled engineering. Notably, we find that a 0.5B LLM yields similar VFM rankings as a 7B LLM while cutting GPU hours by 8x, enabling more efficient evaluation. By offering a comprehensive and transparent benchmark, we hope AVA-Bench lays the foundation for the next generation of VFMs.

BioCLIP 2: Emergent Properties from Scaling Hierarchical Contrastive Learning

Foundation models trained at scale exhibit remarkable emergent behaviors, learning new capabilities beyond their initial training objectives. We find such emergent behaviors in biological vision models via large-scale contrastive vision-language training. To achieve this, we first curate TreeOfLife-200M, comprising 214 million images of living organisms, the largest and most diverse biological organism image dataset to date. We then train BioCLIP 2 on TreeOfLife-200M to distinguish different species. Despite the narrow training objective, BioCLIP 2 yields extraordinary accuracy when applied to various biological visual tasks such as habitat classification and trait prediction. We identify emergent properties in the learned embedding space of BioCLIP 2. At the inter-species level, the embedding distribution of different species aligns closely with functional and ecological meanings (e.g., beak sizes and habitats). At the intra-species level, instead of being diminished, the intra-species variations (e.g., life stages and sexes) are preserved and better separated in subspaces orthogonal to inter-species distinctions. We provide formal proof and analyses to explain why hierarchical supervision and contrastive objectives encourage these emergent properties. Crucially, our results reveal that these properties become increasingly significant with larger-scale training data, leading to a biologically meaningful embedding space.

Revisiting semi-supervised learning in the era of foundation models

Semi-supervised learning (SSL) leverages abundant unlabeled data alongside limited labeled data to enhance learning. As vision foundation models (VFMs) increasingly serve as the backbone of vision applications, it remains unclear how SSL interacts with these pre-trained models. To address this gap, we develop new SSL benchmark datasets where frozen VFMs underperform and systematically evaluate representative SSL methods. We make a surprising observation: parameter-efficient fine-tuning (PEFT) using only labeled data often matches SSL performance, even without leveraging unlabeled data. This motivates us to revisit self-training, a conceptually simple SSL baseline, where we use the supervised PEFT model to pseudo-label unlabeled data for further training. To overcome the notorious issue of noisy pseudo-labels, we propose ensembling multiple PEFT approaches and VFM backbones to produce more robust pseudo-labels. Empirical results validate the effectiveness of this simple yet powerful approach, providing actionable insights into SSL with VFMs and paving the way for more scalable and practical semi-supervised learning in the era of foundation models.

MEDA: Dynamic KV Cache Allocation for Efficient Multimodal Long-Context Inference

Long-context Multimodal Large Language Models (MLLMs) that incorporate long text-image and text-video modalities, demand substantial resources as their multimodal Key-Value (KV) caches grow with increasing input lengths, challenging inference efficiency. Existing methods for KV cache compression, in both text-only and multimodal LLMs, have neglected attention density variations across layers, thus often adopting uniform or progressive reduction strategies for layer-wise cache allocation. In this work, we propose MEDA, a dynamic layer-wise KV cache allocation method for efficient multimodal long-context inference. As its core, MEDA utilizes cross-modal attention entropy to determine the KV cache size at each MLLMs layer. Given the dynamically allocated KV cache size at each layer, MEDA also employs a KV pair selection scheme to identify which KV pairs to select and a KV pair merging strategy that merges the selected and non-selected ones to preserve information from the entire context. MEDA achieves up to 72% KV cache memory reduction and 2.82 times faster decoding speed, while maintaining or enhancing performance on various multimodal tasks in long-context settings, including multi-images and long-video scenarios. Our code is released at https://github.com/AIoT-MLSys-Lab/MEDA.

Finer-CAM: Spotting the Difference Reveals Finer Details for Visual Explanation

Class activation map (CAM) has been widely used to highlight image regions that contribute to class predictions. Despite its simplicity and computational efficiency, CAM often struggles to identify discriminative regions that distinguish visually similar fine-grained classes. Prior efforts address this limitation by introducing more sophisticated explanation processes, but at the cost of extra complexity. In this paper, we propose Finer-CAM, a method that retains CAM's efficiency while achieving precise localization of discriminative regions. Our key insight is that the deficiency of CAM lies not in "how" it explains, but in "what" it explains}. Specifically, previous methods attempt to identify all cues contributing to the target class's logit value, which inadvertently also activates regions predictive of visually similar classes. By explicitly comparing the target class with similar classes and spotting their differences, Finer-CAM suppresses features shared with other classes and emphasizes the unique, discriminative details of the target class. Finer-CAM is easy to implement, compatible with various CAM methods, and can be extended to multi-modal models for accurate localization of specific concepts. Additionally, Finer-CAM allows adjustable comparison strength, enabling users to selectively highlight coarse object contours or fine discriminative details. Quantitatively, we show that masking out the top 5% of activated pixels by Finer-CAM results in a larger relative confidence drop compared to baselines. The source code and demo are available at https://github.com/Imageomics/Finer-CAM.

Prompt-CAM: A Simpler Interpretable Transformer for Fine-Grained Analysis

We present a simple usage of pre-trained Vision Transformers (ViTs) for fine-grained analysis, aiming to identify and localize the traits that distinguish visually similar categories, such as different bird species or dog breeds. Pre-trained ViTs such as DINO have shown remarkable capabilities to extract localized, informative features. However, using saliency maps like Grad-CAM can hardly point out the traits: they often locate the whole object by a blurred, coarse heatmap, not traits. We propose a novel approach Prompt Class Attention Map (Prompt-CAM) to the rescue. Prompt-CAM learns class-specific prompts to a pre-trained ViT and uses the corresponding outputs for classification. To classify an image correctly, the true-class prompt must attend to the unique image patches not seen in other classes' images, i.e., traits. As such, the true class's multi-head attention maps reveal traits and their locations. Implementation-wise, Prompt-CAM is almost a free lunch by simply modifying the prediction head of Visual Prompt Tuning (VPT). This makes Prompt-CAM fairly easy to train and apply, sharply contrasting other interpretable methods that design specific models and training processes. It is even simpler than the recently published INterpretable TRansformer (INTR), whose encoder-decoder architecture prevents it from leveraging pre-trained ViTs. Extensive empirical studies on a dozen datasets from various domains (e.g., birds, fishes, insects, fungi, flowers, food, and cars) validate Prompt-CAM superior interpretation capability.

Lessons Learned from a Unifying Empirical Study of Parameter-Efficient Transfer Learning (PETL) in Visual Recognition

Parameter-efficient transfer learning (PETL) has attracted significant attention lately, due to the increasing size of pre-trained models and the need to fine-tune (FT) them for superior downstream performance. This community-wide enthusiasm has sparked a plethora of new methods. Nevertheless, a systematic study to understand their performance and suitable application scenarios is lacking, leaving questions like when to apply PETL and which method to use largely unanswered. In this paper, we conduct a unifying empirical study of representative PETL methods in the context of Vision Transformers. We systematically tune their hyper-parameters to fairly compare their accuracy on downstream tasks. Our study not only offers a valuable user guide but also unveils several new insights. First, if tuned carefully, different PETL methods can obtain quite similar accuracy in the low-shot benchmark VTAB-1K. This includes simple methods like FT the bias terms that were reported inferior. Second, though with similar accuracy, we find that PETL methods make different mistakes and high-confidence predictions, likely due to their different inductive biases. Such an inconsistency (or complementariness) opens up the opportunity for ensemble methods, and we make preliminary attempts at this. Third, going beyond the commonly used low-shot tasks, we find that PETL is also useful in many-shot regimes -- it achieves comparable and sometimes better accuracy than full FT, using much fewer learnable parameters. Last but not least, we investigate PETL's ability to preserve a pre-trained model's robustness to distribution shifts (e.g., a CLIP backbone). Perhaps not surprisingly, PETL methods outperform full FT alone. However, with weight-space ensembles, the fully FT model can achieve a better balance between downstream and out-of-distribution performance, suggesting a future research direction for PETL.

Fine-Tuning is Fine, if Calibrated

Fine-tuning is arguably the most straightforward way to tailor a pre-trained model (e.g., a foundation model) to downstream applications, but it also comes with the risk of losing valuable knowledge the model had learned in pre-training. For example, fine-tuning a pre-trained classifier capable of recognizing a large number of classes to master a subset of classes at hand is shown to drastically degrade the model's accuracy in the other classes it had previously learned. As such, it is hard to further use the fine-tuned model when it encounters classes beyond the fine-tuning data. In this paper, we systematically dissect the issue, aiming to answer the fundamental question, ''What has been damaged in the fine-tuned model?'' To our surprise, we find that the fine-tuned model neither forgets the relationship among the other classes nor degrades the features to recognize these classes. Instead, the fine-tuned model often produces more discriminative features for these other classes, even if they were missing during fine-tuning! {What really hurts the accuracy is the discrepant logit scales between the fine-tuning classes and the other classes}, implying that a simple post-processing calibration would bring back the pre-trained model's capability and at the same time unveil the feature improvement over all classes. We conduct an extensive empirical study to demonstrate the robustness of our findings and provide preliminary explanations underlying them, suggesting new directions for future theoretical analysis. Our code is available at https://github.com/OSU-MLB/Fine-Tuning-Is-Fine-If-Calibrated.

CompBench: A Comparative Reasoning Benchmark for Multimodal LLMs

The ability to compare objects, scenes, or situations is crucial for effective decision-making and problem-solving in everyday life. For instance, comparing the freshness of apples enables better choices during grocery shopping, while comparing sofa designs helps optimize the aesthetics of our living space. Despite its significance, the comparative capability is largely unexplored in artificial general intelligence (AGI). In this paper, we introduce CompBench, a benchmark designed to evaluate the comparative reasoning capability of multimodal large language models (MLLMs). CompBench mines and pairs images through visually oriented questions covering eight dimensions of relative comparison: visual attribute, existence, state, emotion, temporality, spatiality, quantity, and quality. We curate a collection of around 40K image pairs using metadata from diverse vision datasets and CLIP similarity scores. These image pairs span a broad array of visual domains, including animals, fashion, sports, and both outdoor and indoor scenes. The questions are carefully crafted to discern relative characteristics between two images and are labeled by human annotators for accuracy and relevance. We use CompBench to evaluate recent MLLMs, including GPT-4V(ision), Gemini-Pro, and LLaVA-1.6. Our results reveal notable shortcomings in their comparative abilities. We believe CompBench not only sheds light on these limitations but also establishes a solid foundation for future enhancements in the comparative capability of MLLMs.

Holistic Transfer: Towards Non-Disruptive Fine-Tuning with Partial Target Data

We propose a learning problem involving adapting a pre-trained source model to the target domain for classifying all classes that appeared in the source data, using target data that covers only a partial label space. This problem is practical, as it is unrealistic for the target end-users to collect data for all classes prior to adaptation. However, it has received limited attention in the literature. To shed light on this issue, we construct benchmark datasets and conduct extensive experiments to uncover the inherent challenges. We found a dilemma -- on the one hand, adapting to the new target domain is important to claim better performance; on the other hand, we observe that preserving the classification accuracy of classes missing in the target adaptation data is highly challenging, let alone improving them. To tackle this, we identify two key directions: 1) disentangling domain gradients from classification gradients, and 2) preserving class relationships. We present several effective solutions that maintain the accuracy of the missing classes and enhance the overall performance, establishing solid baselines for holistic transfer of pre-trained models with partial target data.