University of Manchester

NVIDIA has added StormScope EU to its Earth-2 family of open models, a regional version trained on European weather and climate data that uses EUMETSAT satellite observations to predict evolving cloud conditions across 16 Meteosat channels at 10-minute intervals. A global forecast can show a storm taking shape, but decisions about solar farms, shipping routes or insured property depend on what happens locally. At GTC Berlin, the model runs in a live pipeline with Earthmover's data platform, updating as new observations arrive and serving results as map tiles so visitors can explore European predictions without downloading whole datasets.

NVIDIA has added StormScope EU to its Earth-2 family of open models, a regional version trained on European weather and climate data that uses EUMETSAT satellite observations to predict evolving cloud conditions across 16 Meteosat channels at 10-minute intervals. A global forecast can show a storm taking shape, but decisions about solar farms, shipping routes or insured property depend on what happens locally. At GTC Berlin, the model runs in a live pipeline with Earthmover's data platform, updating as new observations arrive and serving results as map tiles so visitors can explore European predictions without downloading whole datasets.

Single-cell RNA sequencing destroys each cell as it measures it and samples only discrete timepoints, so one cell's expression profile cannot be followed continuously. The Latent Cell-State Model approximates cell-state transitions as a Markov chain and adds a low-dimensional continuous latent variable to generate expression counts given cell state, time and gene. On the pancreatic development and Waddington-OT reprogramming datasets, maximum mean discrepancy between generated and real samples at held-out timepoints was 0.011 and 0.006 (pancreatic days 3 and 6) and 0.030 and 0.018 (reprogramming days 3.5 and 4), lower than the three compared methods.

Single-cell RNA sequencing destroys each cell as it measures it and samples only discrete timepoints, so one cell's expression profile cannot be followed continuously. The Latent Cell-State Model approximates cell-state transitions as a Markov chain and adds a low-dimensional continuous latent variable to generate expression counts given cell state, time and gene. On the pancreatic development and Waddington-OT reprogramming datasets, maximum mean discrepancy between generated and real samples at held-out timepoints was 0.011 and 0.006 (pancreatic days 3 and 6) and 0.030 and 0.018 (reprogramming days 3.5 and 4), lower than the three compared methods.

Compressing visual tokens for video large language models costs feature comparisons in the compression decisions themselves. This method aligns reference-frame dependencies, motion vectors and residuals already present in compressed video to the visual token grid, uses them to propose small candidate sets, then verifies with local visual similarity. Across three video LLMs and four video-understanding benchmarks, performance matches the training-free baseline FlashVID and sometimes exceeds it, while comparisons for token protection and temporal matching drop substantially.

Compressing visual tokens for video large language models costs feature comparisons in the compression decisions themselves. This method aligns reference-frame dependencies, motion vectors and residuals already present in compressed video to the visual token grid, uses them to propose small candidate sets, then verifies with local visual similarity. Across three video LLMs and four video-understanding benchmarks, performance matches the training-free baseline FlashVID and sometimes exceeds it, while comparisons for token protection and temporal matching drop substantially.

Using 290,026 outpatient oncology notes from 21,564 patients treated for lung and head-and-neck cancer at a UK specialist oncology centre, the study compared RadBERT and PathologyBERT with two local strategies: OncoNote-RadBERT, produced by continued masked language model pretraining, reached a validation perplexity of 2.10, while OncoNoteBERT, trained from scratch with an oncology-specific WordPiece tokenizer, reached 2.83 but the most efficient tokenisation (subword fertility 1.224, normalised sequence length 0.7636, about 23.6% fewer tokens) and returned clinically acceptable predictions for 12 of 13 masked-token probes versus 7 of 13 for OncoNote-RadBERT; both local models represented the institutional placeholder [REMOVED] as a single learnable token.

Using 290,026 outpatient oncology notes from 21,564 patients treated for lung and head-and-neck cancer at a UK specialist oncology centre, the study compared RadBERT and PathologyBERT with two local strategies: OncoNote-RadBERT, produced by continued masked language model pretraining, reached a validation perplexity of 2.10, while OncoNoteBERT, trained from scratch with an oncology-specific WordPiece tokenizer, reached 2.83 but the most efficient tokenisation (subword fertility 1.224, normalised sequence length 0.7636, about 23.6% fewer tokens) and returned clinically acceptable predictions for 12 of 13 masked-token probes versus 7 of 13 for OncoNote-RadBERT; both local models represented the institutional placeholder [REMOVED] as a single learnable token.

The work proposes DINO-3DRA, a dual-path framework that injects frozen 2D vision foundation model DINOv3-Small features into a trainable 3D U-Net backbone through Room-Lite spatial mixing and calibrated residual fusion, achieving state-of-the-art aneurysm segmentation on multi-centre 3D rotational angiography (3DRA) @neurIST data (233 patients, four institutions) with Dice 0.758, HD95 2.75 mm and a 13% gain over nnU-Net using only 5.72M trainable parameters; ablations attribute the gains to structured cross-dimensional transfer rather than loss design, and without fine-tuning on CADA (n=45) and SHINY-ICARUS (n=30) it reduces Dice<0.5 failures from 11.1% and 3.3% to 0%.

The work proposes DINO-3DRA, a dual-path framework that injects frozen 2D vision foundation model DINOv3-Small features into a trainable 3D U-Net backbone through Room-Lite spatial mixing and calibrated residual fusion, achieving state-of-the-art aneurysm segmentation on multi-centre 3D rotational angiography (3DRA) @neurIST data (233 patients, four institutions) with Dice 0.758, HD95 2.75 mm and a 13% gain over nnU-Net using only 5.72M trainable parameters; ablations attribute the gains to structured cross-dimensional transfer rather than loss design, and without fine-tuning on CADA (n=45) and SHINY-ICARUS (n=30) it reduces Dice<0.5 failures from 11.1% and 3.3% to 0%.

Drawing on Karen Barad's (2007) concepts of intra-action and the agential cut, this chapter employs a poetic inquiry approach situated in three multi-academy trusts in England, and through two agential cuts embodied in distinct poetic forms materialises how place entangles with governance policy, architecture, and actors to configure parental engagement, thereby repositioning place as a co-agentic participant in governance.

Drawing on Karen Barad's (2007) concepts of intra-action and the agential cut, this chapter employs a poetic inquiry approach situated in three multi-academy trusts in England, and through two agential cuts embodied in distinct poetic forms materialises how place entangles with governance policy, architecture, and actors to configure parental engagement, thereby repositioning place as a co-agentic participant in governance.

The work presents VyPER, a geometric learning framework that represents collider events as hypergraphs with a physics-inspired topology, combining supervised classification of hyperedges for assigning measured jets and charged leptons to parent particles with a diffusion model for predicting unmeasured neutrino kinematics, optimized jointly through a joint loss function within a unified framework, and compares its performance to existing analytical and machine-learning-based reconstruction techniques across several proton-proton collision processes, indicating that accurate event reconstruction is achievable across a diverse range of Standard Model physics processes.

The work presents VyPER, a geometric learning framework that represents collider events as hypergraphs with a physics-inspired topology, combining supervised classification of hyperedges for assigning measured jets and charged leptons to parent particles with a diffusion model for predicting unmeasured neutrino kinematics, optimized jointly through a joint loss function within a unified framework, and compares its performance to existing analytical and machine-learning-based reconstruction techniques across several proton-proton collision processes, indicating that accurate event reconstruction is achievable across a diverse range of Standard Model physics processes.