Skip to main content
Search

The University of Texas at Austin

23 related articles

Subscribe to updates
  1. arXiv

    AdamW's fixed points match an implicit objective, but its own dynamics can leave them unstable

    AdamW's fixed points coincide with the stationary points of a constrained, regularized implicit objective, yet those points need not be stable under AdamW's own dynamics. Working in the continuous-time limit, the analysis gives exact one-dimensional stability conditions and sufficient higher-dimensional ones, with examples where even a strongly convex objective loses stability and, in the common momentum regime, trajectories oscillate persistently instead of converging. Rescaling weight decay by the strength of the gradient update yields AdamMe, which keeps the same fixed points and admits a Lyapunov function guaranteeing that bounded trajectories converge.

  2. bioRxiv

    Long-context regulatory models lose their edge across species: at fish distance 20 kb context turns negative while 1 kb local sequence still transfers

    FunFormer compresses up to 20 kb of sequence into a fixed token set for binary regulatory-activity prediction, paired with an architecture- and parameter-matched control that reads only the central 1 kb, and both are transferred among six vertebrates. Within species the 20 kb version has the highest mean AUROC in all six genomes, 1.0 to 3.8 percentage points above three published 1 kb baselines; across species that context gain decays with divergence, nears zero at the amniote split, and reverses between 320 and 435 Mya, while the 1 kb model still transfers into fish.

  3. arXiv

    SAM 3 at 4-bit weights stops grabbing the wrong object: SPQ lifts RefCOCO hit rate from 42.1 to 72.5

    Quantizing SAM 3's image-encoder weights to 4 bits while keeping activations at 8 bits produces prompt-alignment drift: masks look plausible but land on the wrong object or lose the target across video frames. SPQ gives more calibration weight to parameters that affect text-concept scores, constrains feature error along low-variance residual directions, and folds a mean-error correction estimated from paired video rollouts into existing biases. Under W4A8, RefCOCO Pr@0.5 rises from 42.1 for PTQ4SAM to 72.5 while retaining 54.8 COCO mask AP, with no added inference operators.

  4. arXiv

    Seismic velocity inversion drops its handcrafted prior: a diffusion model as projector cuts Marmousi MAE from 94.80 to 92.68

    Full waveform inversion reconstructs subsurface velocity models from seismic records, but it is highly nonlinear and ill-posed, and an inaccurate initial model or noisy data can push it to a wrong solution. Bayesian full waveform inversion built on annealed Stein variational gradient descent previously constrained its particles with handcrafted Gaussian or smoothing priors that cannot represent complex geology. This work embeds a diffusion model trained on the GeoFWI dataset as an implicit geological projection operator inside the particle update, lowering mean absolute error on the Marmousi synthetic model from 94.80 with the classic method to 92.68, raising structural similarity from 0.69 to 0.72, and concentrating posterior uncertainty near deep structures and high-contrast interfaces.

  5. arXiv

    Networked average-reward multi-agent learning gets finite-sample bounds: decentralized updates certify only stationarity, strong entropy restores global optimality

    In networked average-reward multi-agent reinforcement learning, each agent updates its policy using only its k-hop neighborhood, and prior results in this setting were tabular convergence to stationary points. The work first gives two counterexamples: even in a fully cooperative, fully observable, ergodic setting, product local policies can stall at a suboptimal local optimum, and when local policies cannot see the global state a vanishing gradient need not certify unilateral optimality. It then proves a finite-sample stationarity bound for a decentralized actor-critic with linear critics, and shows that strong fixed-state entropy regularization upgrades the guarantee to global optimality.

  6. arXiv

    Marine seismic multiple suppression moves to a self-supervised network: no smoothness constraint, better weak-signal and waveform preservation

    In marine seismic exploration, SRME (surface-related multiple elimination) predicts multiples and then subtracts them from the recorded data with a matching filter; the classical route uses either a fixed-length stationary filter or a non-stationary regression with a local smoothness constraint, which limits it where signal energy changes rapidly. PDL replaces that matching filter with a self-supervised neural network trained under the adaptive matching filtering equation, cutting the seismic profile into many patches so training yields a globally optimal non-stationary match. On synthetic data and the Viking Graben field dataset, PDL leaves fewer multiple residuals than the classical non-stationary least-squares matching filter and better preserves weak primaries and waveform continuity.

  7. arXiv

    Generated meshes no longer need a full redraw: editing only the bad faces lifts quad-equivalent F1 from 0.294 to 0.340

    Generated meshes often carry holes, stray faces and duplicated faces; MeshRepair reads the draft as a prefix and autoregressively emits only a deletion-then-addition edit sequence, raising quad-equivalent face F1 on LATO2 drafts from 0.294 to 0.340 where rule-based repair tools leave it flat or lower.

  8. arXiv

    Speech LLMs stop pooling at a fixed frame rate: a learned boundary policy cuts English audio tokens by about a quarter

    ASSET trains a speech encoder's segment boundaries with reinforcement learning, reaching recognition close to a transcript-assisted character-boundary oracle on LibriSpeech with roughly 25% fewer audio tokens.

  9. arXiv

    Robot rewards stop chasing lighting and camera angle: task-relevant dynamics lift MetaWorld success to 94.3%

    DiTaR builds video rewards on task-relevant latent dynamics, reaching 94.3% final success on MetaWorld-v3 versus 65.2% for the strongest video-reward baseline ReWiND, and raising real-robot offline RL success from 43.3% to 83.3%.

  10. Brain maps now decode to tasks, concepts and domains at once, not just a task name

    Functional decoding of brain activation maps has a new route: NiCLIP aligns neuroimaging article text with coordinate-derived activation maps in one shared latent space through contrastive learning, then uses a cognitive ontology to turn similarity into posterior probabilities over tasks, concepts and domains. Widely used correlation decoders and GC-LDA rely on precomputed meta-analytic maps and fixed vocabularies and give task-level predictions only. On Human Connectome Project group-level maps across seven domains, the best configuration reached 62.86% task recall against a 20.71% maximum for Neurosynth and GC-LDA, while the same model dropped markedly on subject-level maps.