Skip to main content
Search

Sapienza University of Rome

11 related articles

Subscribe to updates
  1. arXiv

    Physics constraints without hard projection: energy-guided transport keeps 100% success on full-PDE 2D Navier-Stokes

    Pretrained flow-matching models learn only the data distribution and often violate conservation laws and boundary conditions at inference; earlier zero-shot fixes used algebraic projections that hit memory bottlenecks and disturb the probability path on dense, full-domain PDEs. Energy-Guided Transport writes the physical residual as a smooth potential energy and applies a continuous dissipative force to intermediate states during integration, obtaining the gradient implicitly through reverse-mode automatic differentiation. Across the heat equation, reaction-diffusion, Burgers' equation and 2D Navier-Stokes, it holds a 100% success rate under full-PDE constraints, generating a 2D Navier-Stokes sample in about 0.94 s where explicit projection fails.

  2. Before diffusion models generalize, they grow spurious attractors absent from training data but repeated in samples

    As the training set grows, diffusion models move from memorizing training samples to generalizing, and between these phases an intermediate kind of attractor appears that matches no training image yet recurs in generated samples. Reading generation as retrieval in a dense associative memory predicts such spurious states once storage capacity is exceeded; on MNIST, FASHION-MNIST, CIFAR-10 and LSUN-CHURCH the spurious fraction rises and then falls at the memorization-to-generalization boundary, with basin volumes between those of memorized and generalized samples.

  3. Screening 2.8 million ordered metals for conductivity: machine learning ranks, electron-phonon calculations check

    Computing metallic conductivity from first principles is expensive, so large chemical spaces have gone largely unscreened. This workflow ranks about 2.8 million ordered compounds using two machine-learning proxies, a BoltzTraP2 conductivity under the constant relaxation-time approximation and the Eliashberg electron-phonon coupling λ, combined by symbolic regression into σSR=0.4σBT/λ, then runs full electron-phonon transport on the candidates. Be2IrPt, H3NaNi, LuIr and LiBePt2 reach room-temperature conductivities of roughly 34 to 38×106 S/m, comparable to aluminum's 36.59×106 S/m but still below Ag, Au and Cu.

  4. Swap one Schelling agent for an LLM and the segregation trajectory holds, but runtime climbs from 2 seconds to 8,917

    In a Mesa implementation of the Schelling segregation model, one agent's symbolic rule for classifying neighbors is replaced by a locally served Qwen3.5 model that classifies through tool calls, while the rest of the dynamics stays fixed. Statistical model checking with MultiVeStA shows the expected ratio of happy agents rising from about 0.65 to about 1.0 within 60 steps in both variants, with point-wise t-tests finding no difference, yet the LLM version takes 8,917 seconds against about 2 seconds. The 0.8b and 2b models show semantic errors, wrong tool-call counts or stuck generation in unit-style checks, while 4b and 9b pass all of them.

  5. Advanced Science

    Gaussian sampling moves into Ising hardware: Falcon signing core cuts dynamic power about fourfold while diffusion images hold up

    Gaussian sampling no longer has to be computed by digital arithmetic: a probabilistic Ising machine produces it from physical stochasticity, and at matched sampling precision the FPGA sampler core uses about 4.0-fold less dynamic power and 4.4-fold fewer logic slices than the isolated Falcon Gaussian core, with diffusion image quality preserved.

  6. arXiv

    Low-rank compression stops spending the same compute per token: training only routers lifts Llama-2-7B downstream accuracy by 7.6 points

    Low-rank compression previously assigned a fixed rank during inference, so every token cost the same compute. LRCC adds one lightweight router per Transformer block that selects among a few nested low-rank paths for each token, training only the routers while the low-rank factors stay frozen. Within the same average active-parameter budget, Llama-2-7B gains 7.6 percentage points in average downstream accuracy over static low-rank compression; at matched batch-size-1 decoding latency, Llama-3.2-1B improves both perplexity and downstream accuracy and Llama-2-7B stays competitive, without specialized CUDA kernels.

  7. arXiv

    AWT uses activation-derived diagonal preconditioning to close 12–60% of the tensor-network compression perplexity gap on Llama, Ministral, and Qwen

    The authors propose Activation-aware Weight Tensorization (AWT), a training-free, solver-agnostic calibration-time preconditioner that applies an activation-derived diagonal equivalent reparameterization to each weight matrix before an unchanged TT/TTN solver and deploys it with only an input-side elementwise rescaling. Across Llama 3.1 8B, Ministral 8B, and Qwen2.5 7B at 2–6 compression, AWT consistently improves vanilla tensorization: it closes 12–35% of the WikiText perplexity gap under single-operator replacement and 27–60% under Llama multi-operator suffix replacement, with gains transferring to HellaSwag and ARC-Challenge.

  8. arXiv

    CCNN built on cup and cap products matches other simplicial complex neural networks on TopoBench

    Addressing the limitation that most topological deep learning approaches, being based on boundary operators and Hodge Laplacians, cannot lift or lower features and signals across more than one dimension per layer, this work proposes the adoption of cup and cap products and formulates the Cup and Cap Topological Neural Network (CCNN), a topological deep learning architecture designed to learn node-based variables (0-cochains) while accounting for many-body interactions such as triangles in the data, validated on the TopoBench datasets where it shows competitiveness with respect to other simplicial complex neural networks.

  9. arXiv

    Labeling agents with their model family split multi-agent groups along the label, raising rounds and tokens and cutting success from 96% to 81%

    Across two no-stakes cooperative games (Leader Election and Exclusion) and the GPQA-Diamond reasoning benchmark, with nine to twenty-five agents drawn from up to five open-weight model families, the study finds that when agents know each other's model family the interaction graph partitions into factions along that label, that the split follows the visible label rather than the underlying architecture (it persists under shuffled labels and neutral color tags and disappears when labels are removed), and that in strictly cooperative tasks labeled groups spend on average 30% more rounds and 55% more tokens to reach a decision while success drops from 96% to 81%.

  10. The added value of echocardiography in pulmonary arterial hypertension risk assessment: an artificial intelligence machine learning-derived analysis of the ULTRA RIGHT VALUE registry

    In a 401-patient multicentre European prospective pulmonary arterial hypertension (PAH) cohort, adding right ventricular echocardiographic parameters—especially right ventricular-pulmonary arterial (RV-PA) coupling indices—to the ESC/ERS and REVEAL 2.0 risk scores improved the c-index for morbi-mortality prediction from 0.73 to 0.78 and from 0.74 to 0.79, respectively (P<0.01), with machine learning identifying RV-PA coupling as the most influential dimension.