LoRA-TTT: Low-Rank Test-Time Training for Vision-Language Models

The rapid advancements in vision-language models (VLMs), such as CLIP, have intensified the need to address distribution shifts between training and testing datasets. Although prior Test-Time Training (TTT) techniques for VLMs have demonstrated robust performance, they predominantly rely on tuning text prompts, a process that demands substantial computational resources and is heavily dependent on entropy-based loss. In this paper, we propose LoRA-TTT, a novel TTT method that leverages Low-Rank Adaptation (LoRA), applied exclusively to the image encoder of VLMs. By introducing LoRA and updating only its parameters during test time, our method offers a simple yet effective TTT approach, retaining the model's initial generalization capability while achieving substantial performance gains with minimal memory and runtime overhead. Additionally, we introduce a highly efficient reconstruction loss tailored for TTT. Our method can adapt to diverse domains by combining these two losses, without increasing memory consumption or runtime. Extensive experiments on two benchmarks, covering 15 datasets, demonstrate that our method improves the zero-shot top-1 accuracy of CLIP-ViT-B/16 by an average of 5.79% on the OOD benchmark and 1.36% on the fine-grained benchmark, efficiently surpassing test-time prompt tuning, without relying on any external models or cache.

Parallel Sequence Modeling via Generalized Spatial Propagation Network

We present the Generalized Spatial Propagation Network (GSPN), a new attention mechanism optimized for vision tasks that inherently captures 2D spatial structures. Existing attention models, including transformers, linear attention, and state-space models like Mamba, process multi-dimensional data as 1D sequences, compromising spatial coherence and efficiency. GSPN overcomes these limitations by directly operating on spatially coherent image data and forming dense pairwise connections through a line-scan approach. Central to GSPN is the Stability-Context Condition, which ensures stable, context-aware propagation across 2D sequences and reduces the effective sequence length to $\sqrt{N}$ for a square map with N elements, significantly enhancing computational efficiency. With learnable, input-dependent weights and no reliance on positional embeddings, GSPN achieves superior spatial fidelity and state-of-the-art performance in vision tasks, including ImageNet classification, class-guided image generation, and text-to-image generation. Notably, GSPN accelerates SD-XL with softmax-attention by over $84\times$ when generating 16K images.

HandsOnVLM: Vision-Language Models for Hand-Object Interaction Prediction

How can we predict future interaction trajectories of human hands in a scene given high-level colloquial task specifications in the form of natural language? In this paper, we extend the classic hand trajectory prediction task to two tasks involving explicit or implicit language queries. Our proposed tasks require extensive understanding of human daily activities and reasoning abilities about what should be happening next given cues from the current scene. We also develop new benchmarks to evaluate the proposed two tasks, Vanilla Hand Prediction (VHP) and Reasoning-Based Hand Prediction (RBHP). We enable solving these tasks by integrating high-level world knowledge and reasoning capabilities of Vision-Language Models (VLMs) with the auto-regressive nature of low-level ego-centric hand trajectories. Our model, HandsOnVLM is a novel VLM that can generate textual responses and produce future hand trajectories through natural-language conversations. Our experiments show that HandsOnVLM outperforms existing task-specific methods and other VLM baselines on proposed tasks, and demonstrates its ability to effectively utilize world knowledge for reasoning about low-level human hand trajectories based on the provided context. Our website contains code and detailed video results \url{https://www.chenbao.tech/handsonvlm/}

Learning to (Learn at Test Time): RNNs with Expressive Hidden States

Self-attention performs well in long context but has quadratic complexity. Existing RNN layers have linear complexity, but their performance in long context is limited by the expressive power of their hidden state. We propose a new class of sequence modeling layers with linear complexity and an expressive hidden state. The key idea is to make the hidden state a machine learning model itself, and the update rule a step of self-supervised learning. Since the hidden state is updated by training even on test sequences, our layers are called Test-Time Training (TTT) layers. We consider two instantiations: TTT-Linear and TTT-MLP, whose hidden state is a linear model and a two-layer MLP respectively. We evaluate our instantiations at the scale of 125M to 1.3B parameters, comparing with a strong Transformer and Mamba, a modern RNN. Both TTT-Linear and TTT-MLP match or exceed the baselines. Similar to Transformer, they can keep reducing perplexity by conditioning on more tokens, while Mamba cannot after 16k context. With preliminary systems optimization, TTT-Linear is already faster than Transformer at 8k context and matches Mamba in wall-clock time. TTT-MLP still faces challenges in memory I/O, but shows larger potential in long context, pointing to a promising direction for future research.

Learning at the Speed of Wireless: Online Real-Time Learning for AI-Enabled MIMO in NextG

Integration of artificial intelligence (AI) and machine learning (ML) into the air interface has been envisioned as a key technology for next-generation (NextG) cellular networks. At the air interface, multiple-input multiple-output (MIMO) and its variants such as multi-user MIMO (MU-MIMO) and massive/full-dimension MIMO have been key enablers across successive generations of cellular networks with evolving complexity and design challenges. Initiating active investigation into leveraging AI/ML tools to address these challenges for MIMO becomes a critical step towards an AI-enabled NextG air interface. At the NextG air interface, the underlying wireless environment will be extremely dynamic with operation adaptations performed on a sub-millisecond basis by MIMO operations such as MU-MIMO scheduling and rank/link adaptation. Given the enormously large number of operation adaptation possibilities, we contend that online real-time AI/ML-based approaches constitute a promising paradigm. To this end, we outline the inherent challenges and offer insights into the design of such online real-time AI/ML-based solutions for MIMO operations. An online real-time AI/ML-based method for MIMO-OFDM channel estimation is then presented, serving as a potential roadmap for developing similar techniques across various MIMO operations in NextG.

Pixel Aligned Language Models

Large language models have achieved great success in recent years, so as their variants in vision. Existing vision-language models can describe images in natural languages, answer visual-related questions, or perform complex reasoning about the image. However, it is yet unclear how localization tasks, such as word grounding or referring localization, can be performed using large language models. In this work, we aim to develop a vision-language model that can take locations, for example, a set of points or boxes, as either inputs or outputs. When taking locations as inputs, the model performs location-conditioned captioning, which generates captions for the indicated object or region. When generating locations as outputs, our model regresses pixel coordinates for each output word generated by the language model, and thus performs dense word grounding. Our model is pre-trained on the Localized Narrative dataset, which contains pixel-word-aligned captioning from human attention. We show our model can be applied to various location-aware vision-language tasks, including referring localization, location-conditioned captioning, and dense object captioning, archiving state-of-the-art performance on RefCOCO and Visual Genome. Project page: https://jerryxu.net/PixelLLM .

IMProv: Inpainting-based Multimodal Prompting for Computer Vision Tasks

In-context learning allows adapting a model to new tasks given a task description at test time. In this paper, we present IMProv - a generative model that is able to in-context learn visual tasks from multimodal prompts. Given a textual description of a visual task (e.g. "Left: input image, Right: foreground segmentation"), a few input-output visual examples, or both, the model in-context learns to solve it for a new test input. We train a masked generative transformer on a new dataset of figures from computer vision papers and their associated captions, together with a captioned large-scale image-text dataset. During inference time, we prompt the model with text and/or image task example(s) and have the model inpaint the corresponding output. We show that training our model with text conditioning and scaling the dataset size improves in-context learning for computer vision tasks by over +10\% AP for Foreground Segmentation, over +5\% gains in AP for Single Object Detection, and almost 20\% lower LPIPS in Colorization. Our empirical results suggest that vision and language prompts are complementary and it is advantageous to use both to achieve better in-context learning performance. Project page is available at https://jerryxu.net/IMProv .

2D-RC: Two-Dimensional Neural Network Approach for OTFS Symbol Detection

Orthogonal time frequency space (OTFS) is a promising modulation scheme for wireless communication in high-mobility scenarios. Recently, a reservoir computing (RC) based approach has been introduced for online subframe-based symbol detection in the OTFS system, where only a limited number of over-the-air (OTA) pilot symbols are utilized for training. However, this approach does not leverage the domain knowledge specific to the OTFS system. This paper introduces a novel two-dimensional RC (2D-RC) method that incorporates the structural knowledge of the OTFS system into the design for online symbol detection on a subframe basis. Specifically, as the channel response acts as a two-dimensional (2D) operation over the transmitted information symbols in the delay-Doppler (DD) domain, the 2D-RC is designed to have a 2D structure to equalize the channel. With the introduced architecture, the 2D-RC can benefit from the predictable channel representation in the DD domain. Moreover, unlike the previous work that requires multiple RCs to learn the channel feature, the 2D-RC only requires a single neural network for detection. Experimental results demonstrate the effectiveness of the 2D-RC approach across different OTFS system variants and modulation orders.

Learning to Estimate: A Real-Time Online Learning Framework for MIMO-OFDM Channel Estimation

In this paper we introduce StructNet-CE, a novel real-time online learning framework for MIMO-OFDM channel estimation, which only utilizes over-the-air (OTA) pilot symbols for online training and converges within one OFDM subframe. The design of StructNet-CE leverages the structure information in the MIMO-OFDM system, including the repetitive structure of modulation constellation and the invariant property of symbol classification to inter-stream interference. The embedded structure information enables StructNet-CE to conduct channel estimation with a binary classification task and accurately learn channel coefficients with as few as two pilot OFDM symbols. Experiments show that the channel estimation performance is significantly improved with the incorporation of structure knowledge. StructNet-CE is compatible and readily applicable to current and future wireless networks, demonstrating the effectiveness and importance of combining machine learning techniques with domain knowledge for wireless communication systems.

Open-Vocabulary Panoptic Segmentation with Text-to-Image Diffusion Models

We present ODISE: Open-vocabulary DIffusion-based panoptic SEgmentation, which unifies pre-trained text-image diffusion and discriminative models to perform open-vocabulary panoptic segmentation. Text-to-image diffusion models have shown the remarkable capability of generating high-quality images with diverse open-vocabulary language descriptions. This demonstrates that their internal representation space is highly correlated with open concepts in the real world. Text-image discriminative models like CLIP, on the other hand, are good at classifying images into open-vocabulary labels. We propose to leverage the frozen representation of both these models to perform panoptic segmentation of any category in the wild. Our approach outperforms the previous state of the art by significant margins on both open-vocabulary panoptic and semantic segmentation tasks. In particular, with COCO training only, our method achieves 23.4 PQ and 30.0 mIoU on the ADE20K dataset, with 8.3 PQ and 7.9 mIoU absolute improvement over the previous state-of-the-art. Project page is available at https://jerryxu.net/ODISE .