Weitian Wang

Rafael Hidalgo, Jesse Parron, Aparna S. Varde et al.

This paper presents a system called Robo-CSK-Organizer that infuses commonsense knowledge from a classical knowledge based to enhance the context recognition capabilities of robots so as to facilitate the organization of detected objects by classifying them in a task-relevant manner. It is particularly useful in multipurpose robotics. Unlike systems relying solely on deep learning tools such as ChatGPT, the Robo-CSK-Organizer system stands out in multiple avenues as follows. It resolves ambiguities well, and maintains consistency in object placement. Moreover, it adapts to diverse task-based classifications. Furthermore, it contributes to explainable AI, hence helping to improve trust and human-robot collaboration. Controlled experiments performed in our work, simulating domestic robotics settings, make Robo-CSK-Organizer demonstrate superior performance while placing objects in contextually relevant locations. This work highlights the capacity of an AI-based system to conduct commonsense-guided decision-making in robotics closer to the thresholds of human cognition. Hence, Robo-CSK-Organizer makes positive impacts on AI and robotics.

Weitian Wang, Lukas Meiner, Rai Shubham et al.

The Visual Geometry Grounded Transformer (VGGT) marks a significant leap forward in 3D scene reconstruction, as it is the first model that directly infers all key 3D attributes (camera poses, depths, and dense geometry) jointly in one pass. However, this joint inference mechanism requires global attention layers that perform all-to-all attention computation on tokens from all views. For reconstruction of large scenes with long-sequence inputs, this causes a significant latency bottleneck. In this paper, we propose head-wise temporal merging (HTTM), a training-free 3D token merging method for accelerating VGGT. Existing merging techniques merge tokens uniformly across different attention heads, resulting in identical tokens in the layers' output, which hinders the model's representational ability. HTTM tackles this problem by merging tokens in multi-head granularity, which preserves the uniqueness of feature tokens after head concatenation. Additionally, this enables HTTM to leverage the spatial locality and temporal correspondence observed at the head level to achieve higher merging ratios with lower merging costs compared to existing methods. Thus, HTTM achieves up to 7x acceleration with negligible performance drops in a GPU-based inference.

Weitian Wang, Rai Shubham, Cecilia De La Parra et al.

In this paper, we propose MixA-Q, a mixed-precision activation quantization framework that leverages intra-layer activation sparsity (a concept widely explored in activation pruning methods) for efficient inference of quantized window-based vision transformers. For a given uniform-bit quantization configuration, MixA-Q separates the batched window computations within Swin blocks and assigns a lower bit width to the activations of less important windows, improving the trade-off between model performance and efficiency. We introduce a Two-Branch Swin Block that processes activations separately in high- and low-bit precision, enabling seamless integration of our method with most quantization-aware training (QAT) and post-training quantization (PTQ) methods, or with simple modifications. Our experimental evaluations over the COCO dataset demonstrate that MixA-Q achieves a training-free 1.35x computational speedup without accuracy loss in PTQ configuration. With QAT, MixA-Q achieves a lossless 1.25x speedup and a 1.53x speedup with only a 1% mAP drop by incorporating activation pruning. Notably, by reducing the quantization error in important regions, our sparsity-aware quantization adaptation improves the mAP of the quantized W4A4 model (with both weights and activations in 4-bit precision) by 0.7%, reducing quantization degradation by 24%.