arXiv Science⌕ Search

arXiv subjects

Hu Cang

Publications and source records attributed to Hu Cang.

6 recordsLinked to original sources

Asking the right question is key to AI discovery in a chain of hypotheses

Artificial intelligence (AI) scientists pursue discovery through a chain of hypotheses, fitting candidate models to existing observations and often choosing experiments where those models disagree most. Yet when early data miss a narrow feature, model selection favors simple curves that agree across the unmea- sured gap, so disagreement-guided search has no reason to look where the new phenomenon lies. When we tested two frontier AI scientists, Claude Sonnet 5.5 and GPT-6.1-Sol, across 32 closed-loop trials with a narrow peak omitted from the starting data, both sampled inside the peak region less often than random search. Surprisingly, the agent that followed model disagreement more closely missed the peak in all 32 trials, whereas discovery succeeded when experiments probed the unmeasured gap early and followed up across the anomaly. Asking the right question, rather than testing only where current models disagree, is therefore key to AI discovery in a chain of hypotheses.

cs.AI↗

More particles or the pair? An exact test of how to improve mode-coupling theory

How a dense liquid arrests into a glass remains a central problem in physics. Mode-coupling theory (MCT), the leading microscopic theory of the glass transition, fails to reproduce mean-field caging. Two routes are proposed to repair it: tracking more-particle correlations, or enforcing repulsion within each colliding pair. Testing both on an exact benchmark liquid, we show that MCT lets touching spheres pass through each other at all orders: adding particles dilutes memory and worsens the transition, whereas restoring pair impenetrability recovers mean-field arrest and the cage size.

cond-mat.soft↗

A Shared Observation Shields Collective Fluctuations while Preserving Local Independence

As a liquid approaches its glass transition, its dynamics turns heterogeneous: mobile and immobile regions coexist, and the four-point susceptibility $χ_4$ that quantifies this heterogeneity grows sharply. Interpreting that growth is subtle, because the collective signals experiments record, such as a tagged particle's trajectory, an overlap function, or a mean field, are generated by the same particles they describe. Here we compute exactly what conditioning on such a shared record does to the population that produced it, for a broad class of stochastically observed systems; the guiding example is a tagged particle and the cage of $z$ neighbors that drives its force history. Using a Girsanov path transformation, we prove that the conditioning multiplies the independent joint law of the $z$ trajectories by exactly one term: a centered-square penalty along the single collective direction the record can see. Any fixed pair of particles stays nearly independent, with covariance falling as $O(z^{-1})$ and mutual information as $O(z^{-2})$, the property known as propagation of chaos, yet the $z(z-1)$ weak pair correlations add coherently into a finite suppression of collective fluctuations, the Schur shield $D - C = -C^2(aI + C)^{-1} \preceq 0$. An exactly solvable Brownian model calibrates the construction. The physical consequence is a calculable baseline for dynamical heterogeneity: conditioning itself contributes a computable, nonpositive amount to the susceptibility of a conditioned ensemble, so the genuine cooperative signal is the excess of the measured $χ_4$ over this baseline rather than over zero, a comparison that existing simulation data can already perform.

cond-mat.stat-mech↗

Particle Contacts Generate Fractional Density Relaxation

Dense-liquid relaxation evolves from local particle collisions to cooperative structural rearrangements. While hard-sphere kinetics determines an early $t^{3/2}$ fractional decay in density correlation functions, collective theories describe the subsequent structural relaxation. A central open question has been how short-time contact physics supplies an exact starting point for the memory kernel governing later times without being modified by subsequent many-body rearrangements. Here we resolve this problem for a broad class of reversible Brownian systems. We prove that hard particle contacts act as reflecting boundaries in configuration space, uniquely dictating the amplitude of the leading $t^{3/2}$ density relaxation. Mechanistically, diffusion samples a contact boundary layer of thickness $O(\sqrt{t})$, which combines with the local density response to produce the fractional signal. We derive an explicit surface formula expressing this amplitude in terms of equilibrium contact probability, normal mobility, and density sensitivity. For monodisperse hard spheres, this yields an exact, fit-free prediction determined entirely by static structure $S(k)$, radial contact value $g(σ^+)$, and short-time diffusion $D_0$. Extending the construction, we determine the corresponding normalization for soft interfaces and prove via a Gram--Schur projection hierarchy that regular collective variables leave the leading contact amplitude strictly invariant. The resulting formulation connects microscopic collision kinetics directly to caging and glass-like structural relaxation, providing an exact microscopic boundary condition for scattering experiments, molecular simulations, and memory-kernel reconstructions.

cond-mat.soft↗

Universality and Falsifiability of Quantum Spacetime Decoherence: A Gauge-Invariant Framework for Gravitational-Wave Phase Diffusion

We develop a fully gauge-invariant and rigorously derived framework for computing the cumulative decoherence of gravitational waves (GWs) propagating through a stochastic quantum spacetime. Working directly with the Riemann-tensor two-point function and exploiting the extreme adiabaticity of cosmological GW propagation, we show that phase diffusion, rather than amplitude attenuation or mode mixing, is the unique leading-order imprint of microscopic curvature fluctuations. Our main theoretical result is a universality theorem: for any quantum-gravity model whose curvature fluctuations possess a finite correlation length, the accumulated phase variance grows linearly with distance, independent of the underlying microphysics. This diffusive scaling contrasts sharply with coherent astrophysical effects and with nonlocal models. The frequency exponent therefore becomes a clean spectral discriminator, separating string-foam recoil, holographic or scale-invariant noise, and causal-set discreteness. We obtain these results from first principles by evaluating the projected Riemann correlator along null geodesics and determining the exact conditions under which deviations from universality can arise. Finally, we outline a hierarchical Bayesian strategy for measuring this effect with LIGO, LISA, and Pulsar Timing Arrays. Although standard Planck-scale fluctuations remain far below current sensitivity, this framework provides a sharp and falsifiable test of exotic quantum-spacetime scenarios, particularly those with macroscopic correlation lengths or strong energy dependence.

gr-qc↗

CRISPR/Cas9 For Photoactivated Localization Microscopy (PALM)

We demonstrate that endonuclease deficient Clustered Regularly Interspaced Short Palindromic Repeats CRISPR-associated Cas9 protein (dCas9) fused to the photo-convertible fluorescence protein monomeric mEos3.1 (dCas9-mEos3) can be used to resolve sub-diffraction limited features of repetitive gene elements, thus providing a new route to investigate high-order chromatin organization at these sites.

q-bio.SC↗