Skip to main content
Search

Institute of Science Tokyo

8 related articles

Subscribe to updates
  1. bioRxiv

    PDZ-binding peptide design moves into a shared latent space: one trained model searches off-target and dual-target objectives without retraining

    PDZ-binding motifs (PBMs) are short C-terminal sequences that determine which PDZ domains a protein binds. Designing PBMs by affinity has generally required a dedicated model per domain and made off-target control hard. This work encodes PDZ and PBM sequences into one 128-dimensional unit hypersphere, predicts pKd from their cosine similarity, and writes single-target, target-minus-off-target and dual-target-mean objectives as directions on that sphere; in four-position redesign for DLG4-PDZ3, 84 of the top 100 candidates fell in the top 1% of measured fitness in a published mutant library, versus 64 for the comparison method.

  2. arXiv

    Same Login Screen, Opposite Labels: LGWM Checks the Transition in 17 ms, Not the Screen

    For GUI agents, the safety signal may lie not in the screen but in the transition that produced it: the same credential screen can be a legitimate result of tapping Sign in or a hijacked result of tapping View order. LGWM skips rendering the next screen and predicts its representation directly in the observation encoder's space, reducing verification to a vector comparison; on RSWT-Bench, where each credential screen appears pixel-identically under both labels, the training-free consistency score reaches 0.987 AUC at 17 ms per decision, on par with the strongest closed-source vision-language models and about ten AUC points above generative GUI world models.

  3. arXiv

    AI agents vote down their own run: a diamond-spin lab rejects a passing experiment, then earns the evidence

    A laboratory of large-language-model agents, each holding a separate research role, asked how many calibration measurements are needed to predict a diamond nitrogen-vacancy spin ensemble's response to arbitrary phase-controlled microwave pulse sequences. The principal-investigator agent tightened a proposed bound of 91 calibrations to an exact minimum of 10; the first acquisition met all ten precommitted accuracy criteria but failed phase and reference controls and was ruled inconclusive, and after the agents revised the protocol without relaxing the criteria, the final measurement supported the prediction while two interpolation methods predicted zero.

  4. arXiv

    SRSP trains a style planner with a frozen TTS model's speech-likelihood reward, beating captioning baselines that see the target audio on five acoustic metrics

    The work first shows that speech-text alignment only weakly predicts downstream acoustic similarity, then proposes Speech-Rewarded Style Planning (SRSP), which trains a text-based style planner with GRPO using the teacher-forced likelihood of target speech tokens under a frozen TTS model as reward; on an English subset of the ISCSLP 2026 CoT-TTS corpus, SRSP outperforms the Base LLM and target-audio-informed captioning baselines on five downstream acoustic metrics and gains in LLM-judged contextual appropriateness and reference consistency.

  5. arXiv

    Three-sensor machine-learning phase-amplitude reduction speeds airfoil wake synchronization eightfold while curbing lift fluctuations

    This study combines phase-amplitude reduction with nonlinear machine-learning sparse-sensor reconstruction to estimate phase and amplitude sensitivity fields from only three airfoil-surface sensors and analytically derive optimal actuation waveforms for fast synchronization of post-stall airfoil wake shedding; the identified optimal forcing alters wake frequency eight times faster than standard sinusoidal actuation, and the amplitude-penalized waveform suppresses lift-coefficient fluctuation relative to the optimal forcing without amplitude penalty.

  6. Hasegawa and Ohzeki compare Ising and QUBO encodings under fixed model, sampler, and step size, finding QUBO has lower Fisher spectral entropy and slower SGD convergence while natural gradient descent makes the encodings converge alike

    Under a controlled protocol that fixes the Boltzmann machine model, simulated-annealing sampler, and learning-rate design, the study compares Ising ({−1,+1}) and QUBO ({0,1}) variable encodings, exploits the identity that the Fisher information matrix equals the covariance of sufficient statistics to visualize empirical moments, and finds that QUBO induces larger cross terms between first- and second-order statistics, creating more small-eigenvalue directions and lowering spectral entropy, which explains slower convergence under stochastic gradient descent, whereas natural gradient descent, which rescales updates by the Fisher information matrix metric, achieves similar convergence across encodings due to reparameterization invariance.

  7. arXiv

    Unite-Audio jointly trains audio representation learning with latent flow matching, reaching competitive text-to-audio generation with a compact model

    The work introduces Unite-Audio, which jointly learns continuous audio representations and latent flow matching for text-to-audio generation; by coupling reconstruction with self-supervised generative prediction it lets the generative objective directly shape the latent space, and it applies Flow-GRPO post-training to improve text-conditioned generation; experiments show competitive text-to-audio performance with a compact latent flow model, and ablation studies confirm the benefit of jointly learning the audio representation and the generative model.

  8. medRxiv

    Diagnostic Value of Large Language Model-Extracted Gross Brain Findings in Neurodegenerative Diseases

    Using 5,613 autopsy cases from the Mayo Clinic Brain Bank collected between 1998 and 2023, this study fine-tuned a large language model to convert narrative gross descriptions into semi-quantitative scores for 39 features (extraction accuracy 0.95 on 200 manually annotated feature-level test examples), then classified seven neuropathologic diagnostic categories with a CatBoost classifier and a second fine-tuned LLM, both including age at death, sex, and brain weight; on a held-out test set of 562 cases CatBoost reached accuracy 0.73, kappa 0.65, and macro-average AUC 0.92, while the text-based LLM reached accuracy 0.75 and kappa 0.68, with macro-average sensitivity 0.66 for both, PSP sensitivity 0.92 and 0.93, MSA sensitivity 0.87 and 0.92, but AD-LBD sensitivity only 0.21 and 0.

You’re all caught up