Skip to main content
Search

Brown University

12 related articles

Subscribe to updates
  1. arXiv

    Agents must now identify the PDE themselves: RSI-PDEbench tests them on 1,001 cases, lifting mean pass rates by up to 181.9% across four models

    Recovering a partial differential equation from sampled solution data demands expert work in equation discovery, coefficient estimation and numerical solver configuration. RSI-PDEbench turns that workflow into a benchmark of 1,001 cases over 42 equation templates, split into noise-free, 50% Gaussian-noise and unseen-template-with-shifted-coefficient settings; the accompanying RSI-PDE agent shortlists candidate equations by residual, refines coefficients and solvers with relative feedback, and keeps improvements behind a validation gate. Across four frontier models its mean pass rate rises 106.5–181.9% over single-run baselines and 3.0–19.6% over native agent harnesses.

  2. arXiv

    Why X posts stall then surge: a fractional physics-informed network splits 24 hours into four windows and lifts the memory order from 0.23 to 0.99

    Models of social media spread have either prescribed a fixed growth function or assumed a fixed way in which past activity fades. A fractional physics-informed neural network writes each post's engagement rate as a per-post viral potential times one shared growth law, with a learnable fractional derivative setting how strongly history matters. On 348 training and 86 test posts from X, splitting the 24-hour window into four separately trained segments let the model reconstruct 64% of held-out trajectories within a mean squared error of 0.01, against 44% for Bass diffusion, the strongest of nine classical baselines.

  3. arXiv

    Alignment without scale: good views let a 109M text JEPA beat 72B language models on the Platonic representation benchmark

    The Platonic Representation Hypothesis, which holds that separately trained models converge on a shared picture of the world, becomes a theorem for JEPAs (joint-embedding predictive architectures): if each modality's positive pairs link the same world states, the encoders agree up to one rotation. The resulting good-view versus bad-view distinction predicts alignment, and a 109M-parameter text JEPA reaches .184 on the PRH benchmark, above every language model evaluated, up to Qwen2.5-72B at .154.

  4. arXiv

    World models stop training a controller separately: LeWAM folds action prediction into one pass, lifting contact-rich success from 28.6% to 89.7%

    LeWAM trains visual representation, action-conditioned latent dynamics, and a stochastic policy in a single stage, raising planning success from 28.6% to 89.7% on contact-rich tasks and from 10.9% to 46.2% on long-horizon tasks while planning 32.3 times faster.

  5. arXiv

    DSReg provably recovers individual world latents without reconstruction, improving sparse control and editing across synthetic, pixel-trained, and external-renderer benchmarks

    The work introduces DSReg (Dependency-Sparsity Regularization), which, on top of the linear-identifiability premise provided by LeJEPA, proves that under Structural Diversity—distinct dependency footprints of latents on observations—each world latent can be recovered up to signed permutation with no reconstruction, no decoder, and no labels; the method applies post hoc to any linearly identified representation, reuses trained checkpoints, and across synthetic regimes, world-model probes, pixel-trained encoders, and external renderers preserves dense prediction while improving individual-latent recovery and downstream sparse control, editing, and few-shot readout.

  6. arXiv

    ZeroMAG generates multimodal adapters zero-shot from unlabeled calibration signals, raising balanced accuracy of frozen EEG foundation models by 7.22 points on average across six held-out datasets

    ZeroMAG introduces a zero-shot multimodal adapter generation framework that, with a frozen EEG encoder and prediction head, uses only unlabeled target calibration recordings and task context to generate adapter weights through a configuration-invariant 1+N adapter, a modality–subject–task condition, and a function-constrained latent space learned from source adapters; across six held-out target datasets and three EEG foundation model backbones it improves balanced accuracy by 7.22 percentage points over EEG-only inference and 4.89 points over direct weight regression, coming within 0.50 points of supervised multimodal adaptation on average, with no target labels or target-side optimization.

  7. arXiv

    Rolling-WAM spreads joint denoising across replanning cycles, reaching 98.1% on LIBERO and 93.3% on RoboTwin 2.0 with a 4.5x steady-state replanning speedup over standard joint WAMs

    Rolling-WAM introduces a rolling world action model that keeps a sliding window of video-action chunks at staggered noise levels, fully denoising only the imminent action chunk each replanning cycle while partially refining farther-future chunks, thereby distributing joint denoising computation across control cycles; it achieves competitive manipulation success on LIBERO, RoboTwin 2.0, and real-world Unitree G1 humanoid tasks while delivering a 4.5x steady-state replanning speedup over standard joint WAMs.

  8. arXiv

    FuseReg trains decoders and DiTs on random encoder-layer subsets, letting one decoder reconstruct from full, sparse, and single-layer fusions and lowering gFID with the generator unchanged

    The work introduces FuseReg, which replaces heuristic layer selection in representation autoencoders (RAEs) with training over random subsets of encoder layers: on ImageNet-256 with a frozen DINOv3-L, a single FuseReg decoder reconstructs from full, sparse, and single-layer fusions without retraining and achieves higher PSNR than decoders specialized to fixed fusions, while decoder replacement alone (with an unchanged RAEv2 DiT-XL generator) reduces unguided gFID and joint regularization of decoder and generator further reduces unguided gFID on DiT-Base.

  9. MergeRepair merges multiple code-task adapters and lifts automated program repair by 2.38% pass@1 and 4.01% pass@10 on StarCoder2 without extra training

    The study introduces MergeRepair, which trains one LoRA adapter per task on the Python split of CommitPackFT (59,113 samples, with Bug Fixes/APR at 19.02%) for Development, Bug Fixes, Misc, Test & QA, and Improvement, merges them with weight-averaging, ties, and dare-ties under equal weights, and proposes continual merging that adds adapters sequentially; evaluated on StarCoder2 and Granite 3B code models with HumanEvalFix pass@1/pass@10, it finds that merging task-specific adapters can improve APR without additional training, reaching up to 2.38% pass@1 and 4.

  10. Nanoscale

    Programmable nanoprobes for molecular imaging of cancer: toward adaptive and context-responsive diagnostics

    This review surveys the design principles and functional architectures of stimuli-responsive nanoprobes, noting that they can achieve context-responsive activation and signal modulation in response to endogenous tumor cues (acidic pH around 6.5-6.8, glutathione at 2-10 mM, enzymatic overexpression, hypoxia below 2% O2, and redox gradients) and exogenous triggers (near-infrared light at 700-1000 nm, magnetic fields, and ultrasound), yielding 5-20-fold signal amplification, and that they can be integrated with multimodal imaging and artificial intelligence for real-time data interpretation and adaptive diagnostics.