News | Intelligent Systems Lab @ PITThttps://pittisl.github.io/news/News from the Intelligent Systems Lab @ PITTen-usThu, 24 Sep 2026 00:00:00 +0000Our paper, Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models, has been …https://pittisl.github.io/publication/2026-vlm-latent-shaping/Thu, 24 Sep 2026 00:00:00 +0000https://pittisl.github.io/news/#2026-09-24-d408de9cOur paper, <a href="https://pittisl.github.io/publication/2026-vlm-latent-shaping/">Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models</a>, has been accepted to <strong>NeurIPS 2026</strong> as a <strong>spotlight presentation</strong>.Our paper, SpatialMind: Spatially Aware On-Device Embodied AI via Viewpoint Integration, has been accepted for …https://pittisl.github.io/publication/2026-spatialmind/Sat, 15 Aug 2026 00:00:00 +0000https://pittisl.github.io/news/#2026-08-15-d73fc794Our paper, <a href="https://pittisl.github.io/publication/2026-spatialmind/">SpatialMind: Spatially Aware On-Device Embodied AI via Viewpoint Integration</a>, has been accepted for publication at the ACM International Conference on Mobile Computing and Networking (MobiCom 2026).Our paper, Reasoning Path and Latent State Analysis for Multi-view Visual Spatial Reasoning: A Cognitive Science …https://pittisl.github.io/publication/2025-remindview-bench/Wed, 17 Jun 2026 00:00:00 +0000https://pittisl.github.io/news/#2026-06-17-1d998688Our paper, <a href="https://pittisl.github.io/publication/2025-remindview-bench/">Reasoning Path and Latent State Analysis for Multi-view Visual Spatial Reasoning: A Cognitive Science Perspective</a>, has been accepted for publication at the European Conference on Computer Vision (ECCV 2026).Our paper, Attribution-based Sparse Activation in Large Language Models, has been accepted for publication at the Ninth …https://pittisl.github.io/publication/2026-sparse-activation-slm/Fri, 01 May 2026 00:00:00 +0000https://pittisl.github.io/news/#2026-05-01-b3f531e7Our paper, <a href="https://pittisl.github.io/publication/2026-sparse-activation-slm/">Attribution-based Sparse Activation in Large Language Models</a>, has been accepted for publication at the Ninth Conference on Machine Learning and Systems (MLSys 2026).Our paper, InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene Complexity, has been …https://pittisl.github.io/publication/2025-infinibench/Sun, 01 Mar 2026 00:00:00 +0000https://pittisl.github.io/news/#2026-03-01-f4583ca3Our paper, <a href="https://pittisl.github.io/publication/2025-infinibench/">InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene Complexity</a>, has been accepted to <strong>CVPR 2026</strong> as an <strong>oral presentation</strong>.Three of our recent papers, InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene …https://pittisl.github.io/publication/2025-infinibench/Mon, 01 Dec 2025 00:00:00 +0000https://pittisl.github.io/news/#2025-12-01-cdcf6234Three of our recent papers, <a href="https://pittisl.github.io/publication/2025-infinibench/">InfiniBench: Infinite Benchmarking for Visual Spatial Reasoning with Customizable Scene Complexity</a>, <a href="https://pittisl.github.io/publication/2025-remindview-bench/">Reasoning Path and Latent State Analysis for Multi-view Visual Spatial Reasoning: A Cognitive Science Perspective</a>, and <a href="https://pittisl.github.io/publication/2025-spatial-reasoning-survey/">Spatial Reasoning in Multimodal Large Language Models: A Survey of Tasks, Benchmarks and Methods</a>, are now available on arXiv.Our paper, MMBERT: Scaled Mixture-of-Experts Multimodal BERT for Robust Chinese Hate Speech Detection under Cloaking …https://pittisl.github.io/publication/2025-mmbert/Sat, 08 Nov 2025 00:00:00 +0000https://pittisl.github.io/news/#2025-11-08-9afaabeeOur paper, <a href="https://pittisl.github.io/publication/2025-mmbert/">MMBERT: Scaled Mixture-of-Experts Multimodal BERT for Robust Chinese Hate Speech Detection under Cloaking Perturbations</a>, has been accepted for publication at the 40th Annual AAAI Conference on Artificial Intelligence (AAAI 2026).Our paper, ProGait: A Multi-Purpose Video Dataset and Benchmark for Transfemoral Prosthesis Users, has been accepted for …https://pittisl.github.io/publication/2025-progait/Tue, 01 Jul 2025 00:00:00 +0000https://pittisl.github.io/news/#2025-07-01-f995a231Our paper, <a href="https://pittisl.github.io/publication/2025-progait/">ProGait: A Multi-Purpose Video Dataset and Benchmark for Transfemoral Prosthesis Users</a>, has been accepted for publication at 2025 International Conference on Computer Vision (ICCV 2025).Our paper, Data Can Speak for Itself: Quality-guided Utilization of Wireless Synthetic Data, has been accepted for …https://pittisl.github.io/publication/2025-syncheck/Sun, 01 Jun 2025 00:00:00 +0000https://pittisl.github.io/news/#2025-06-01-836e96aeOur paper, <a href="https://pittisl.github.io/publication/2025-syncheck/">Data Can Speak for Itself: Quality-guided Utilization of Wireless Synthetic Data</a>, has been accepted for publication at the ACM International Conference on Mobile Systems, Applications, and Services (MobiSys 2025) with the <strong>best paper award!</strong>Our paper, Never Start from Scratch: Expediting On-Device LLM Personalization via Explainable Model Selection, has been …https://pittisl.github.io/publication/2025-xpert/Sun, 01 Jun 2025 00:00:00 +0000https://pittisl.github.io/news/#2025-06-01-3a101ac2Our paper, <a href="https://pittisl.github.io/publication/2025-xpert/">Never Start from Scratch: Expediting On-Device LLM Personalization via Explainable Model Selection</a>, has been accepted for publication at the ACM International Conference on Mobile Systems, Applications, and Services (MobiSys 2025).Our paper, PhyT2V: LLM-Guided Iterative Self-Refinement for Physics-Grounded Text-to-Video Generation, has been accepted …https://pittisl.github.io/publication/2024-phyt2v/Tue, 01 Apr 2025 00:00:00 +0000https://pittisl.github.io/news/#2025-04-01-02d8c154Our paper, <a href="https://pittisl.github.io/publication/2024-phyt2v/">PhyT2V: LLM-Guided Iterative Self-Refinement for Physics-Grounded Text-to-Video Generation</a>, has been accepted for publication at the Conference on Computer Vision and Pattern Recognition 2025 (CVPR 2025).Our paper, Tackling Intertwined Data and Device Heterogeneities in Federated Learning with Unlimited Staleness, has been …https://pittisl.github.io/publication/2023-intertwined-heterogeneity/Sun, 01 Dec 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-12-01-d70639aeOur paper, <a href="https://pittisl.github.io/publication/2023-intertwined-heterogeneity/">Tackling Intertwined Data and Device Heterogeneities in Federated Learning with Unlimited Staleness</a>, has been accepted for publication at the 39th Annual Conference on Artificial Intelligence (AAAI 2025).Two of our papers, When Device Delays Meet Data Heterogeneity in Federated AIoT Applications and Modality Plug-and-Play: …https://pittisl.github.io/publication/2025-aiot/Sun, 01 Dec 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-12-01-2d99945cTwo of our papers, <a href="https://pittisl.github.io/publication/2025-aiot/">When Device Delays Meet Data Heterogeneity in Federated AIoT Applications</a> and <a href="https://pittisl.github.io/publication/2023-mpnp-llm/">Modality Plug-and-Play: Runtime Modality Adaptation in LLM-Driven Autonomous mobile Systems</a>, have been accepted for publication at the 2025 ACM International Conference on Mobile Computing and Networking (MobiCom'25).The preprint of our recent work on inference-time text-to-video (T2V) generation refinement, PhyT2V: LLM-Guided …https://pittisl.github.io/publication/2024-phyt2v/Fri, 01 Nov 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-11-01-c6af6499The preprint of our recent work on inference-time text-to-video (T2V) generation refinement, <a href="https://pittisl.github.io/publication/2024-phyt2v/">PhyT2V: LLM-Guided Iterative Self-Refinement for Physics-Grounded Text-to-Video Generation</a>, is now available on ArXiv.Our paper, Perceptual-Centric Image Super-Resolution using Heterogeneous Processors on Mobile Devices, has been accepted …https://pittisl.github.io/publication/2024-fye-sr/Sun, 01 Sep 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-09-01-125ddec0Our paper, <a href="https://pittisl.github.io/publication/2024-fye-sr/">Perceptual-Centric Image Super-Resolution using Heterogeneous Processors on Mobile Devices</a>, has been accepted for publication at the 2024 ACM International Conference on Mobile Computing and Networking (MobiCom'24).The preprint of our recent work on preventing illegal model adaptation, FreezeAsGuard: Mitigating Illegal Adaptation of …https://pittisl.github.io/publication/2024-freezeasguard/Sat, 01 Jun 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-06-01-dd8d0672The preprint of our recent work on preventing illegal model adaptation, <a href="https://pittisl.github.io/publication/2024-freezeasguard/">FreezeAsGuard: Mitigating Illegal Adaptation of Diffusion Models via Selective Tensor Freezing</a>, and research on the sparsification of Small Language Models (SLMs), <a href="https://pittisl.github.io/publication/2026-sparse-activation-slm/">Achieving Sparse Activation in Small Language Models</a>, are now available on arXiv.We are happy to publish the dataset of human airway measurements, produced by our integrated AI and sensing systems for …https://pittisl.github.io/dataset/#awareMon, 01 Jan 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-01-01-9e957b04We are happy to publish the dataset of human airway measurements, produced by our integrated AI and sensing systems for smart pulmonary telemedicine, namely <a href="https://pittisl.github.io/dataset/#aware">Acoustic Waveform Respiratory Evaluation (AWARE)</a>.Our paper, Towards Green AI in Fine-Tuning Large Language Models via Adaptive Backpropagation, has been accepted for …https://pittisl.github.io/publication/2023-greentrainer/Mon, 01 Jan 2024 00:00:00 +0000https://pittisl.github.io/news/#2024-01-01-3b3ef521Our paper, <a href="https://pittisl.github.io/publication/2023-greentrainer/">Towards Green AI in Fine-Tuning Large Language Models via Adaptive Backpropagation</a>, has been accepted for publication at the 2024 International Conference on Learning Representations (ICLR).The preprint of our recent work on runtime modality adaptation for embodied AI, Modality Plug-and-Play: Elastic Modality …https://pittisl.github.io/publication/2023-mpnp-llm/Fri, 01 Dec 2023 00:00:00 +0000https://pittisl.github.io/news/#2023-12-01-9fc170c4The preprint of our recent work on runtime modality adaptation for embodied AI, <a href="https://pittisl.github.io/publication/2023-mpnp-llm/">Modality Plug-and-Play: Elastic Modality Adaptation in Multimodal LLMs for Embodied AI</a>, has been made publicly available <a href="https://arxiv.org/abs/2312.07886" target="_blank" rel="noopener">on arXiv</a>.Two of our recent works, Towards Green AI in Fine-tuning Large Language Models via Adaptive Backpropagation …https://pittisl.github.io/publication/2023-greentrainer/Fri, 01 Sep 2023 00:00:00 +0000https://pittisl.github.io/news/#2023-09-01-1b9e0d64Two of our recent works, <a href="https://pittisl.github.io/publication/2023-greentrainer/">Towards Green AI in Fine-tuning Large Language Models via Adaptive Backpropagation</a> (“GreenTrainer”) and <a href="https://pittisl.github.io/publication/2023-intertwined-heterogeneity/">Tackling the Unlimited Staleness in Federated Learning with Intertwined Data and Device Heterogeneities</a>, are now available online on arXiv.Two papers from our lab, ElasticTrainer: Speeding Up On-Device Training with Runtime Elastic Tensor Selection and …https://pittisl.github.io/publication/2023-elastictrainer/Thu, 01 Jun 2023 00:00:00 +0000https://pittisl.github.io/news/#2023-06-01-9ead7393Two papers from our lab, <a href="https://pittisl.github.io/publication/2023-elastictrainer/">ElasticTrainer: Speeding Up On-Device Training with Runtime Elastic Tensor Selection</a> and <a href="https://pittisl.github.io/publication/2023-ptease/">PTEase: Objective Airway Examination for Pulmonary Telemedicine using Commodity Smartphones</a>, are accepted and presented at MobiSys 2023.AiFi: AI-Enabled WiFi Interference Cancellation with Commodity PHY-Layer Information is accepted and presented at SenSys …https://pittisl.github.io/publication/2022-aifi/Tue, 01 Nov 2022 00:00:00 +0000https://pittisl.github.io/news/#2022-11-01-d95753eb<a href="https://pittisl.github.io/publication/2022-aifi/">AiFi: AI-Enabled WiFi Interference Cancellation with Commodity PHY-Layer Information</a> is accepted and presented at SenSys 2022.Our AgileNN work is accepted and presented at MobiCom 2022. The published paper, Real-time Neural Network Inference on …https://pittisl.github.io/publication/2022-agilenn/Sat, 01 Oct 2022 00:00:00 +0000https://pittisl.github.io/news/#2022-10-01-ae2e85d7Our AgileNN work is accepted and presented at MobiCom 2022. The published paper, <a href="https://pittisl.github.io/publication/2022-agilenn/">Real-time Neural Network Inference on Extremely Weak Devices: Agile Offloading with Explainable AI</a>, is also available online.