arXiv subjects

Corey O'Hern

Publications and source records attributed to Corey O'Hern.

3 recordsLinked to original sources

arXiv2026

How Good Are Frontier Models at Physics? Expert Re-Grading Reveals Broken Evaluations and Near-Saturation of Leading Benchmarks

Low reported scores on leading physics benchmarks, including those featured in the Artificial Analysis Intelligence Index (2026), suggest that frontier language models still struggle with advanced physics, a demanding test of their scientific reasoning and quantitative problem-solving abilities. Yet this impression does not always align with domain experts' experiences using these models in their work. We revisit these reported findings by evaluating frontier models on six widely used physics benchmarks and auditing them with experts, focusing on text-only problems with verifiable final answers. For each subfield of physics, faculty and graduate researchers with relevant expertise carefully review problem statements, reference solutions, and model responses to distinguish genuine model errors from grader errors, incorrect reference solutions, and ambiguous or underspecified questions. Most audited cases initially evaluated as incorrect reflect these benchmarking issues rather than errors in the models' physics reasoning. We then ask experts to address these benchmarking issues by correcting erroneous reference solutions and repairing or excluding flawed questions. We find that GPT-5.6-Sol's measured mean@4 rises from 47.3% to 78.7% on HLE-Physics and from 61.0% to 87.2% on CMT-Benchmark, while its corrected pass@4 reaches 94.4% on the 54 retained CritPt challenges. Corrected scores are computed on the retained evaluation subsets following expert review. Scores on the audited subsets of UGPhysics, PRISM-Physics, and PHYBench also rise substantially after correction. These findings suggest that current benchmarks substantially understate frontier models' ability to solve well-posed physics problems. Near-saturation on these closed-ended tasks highlights the need for more demanding, expert-validated evaluations.

cs.AI↗

arXiv2016

Granular Impact: A Grain-scale Approach

This work summarizes a series of studies on two-dimensional granular impact, where an intruding object strikes a granular material at high speed. Many previous studies on granular impact have used a macroscopic force law, which is dominated by an inertial drag term proportional to the intruder velocity squared. The primary focus here is on the microscopic force response of the granular material, and how the grain-scale effects give rise to this inertial drag term. We show that the inertial drag arises from intermittent collisions with force-chain-like structures. We construct a simple collisional model to explain the inertial drag, as well as off-axis instability and rotations. Finally, we show how the granular response changes when the intruder speed approaches $d/t_c$, leading to a failure of the inertial drag description in this regime. Here, $d$ is the mean particle diameter and $t_c$ the characteristic momentum-transfer time between two grains.

cond-mat.soft↗

arXiv2009

Comparison of low amplitude oscillatory shear in experimental and computational studies of model foams

A fundamental difference between fluids and solids is their response to applied shear. Solids possess static shear moduli, while fluids do not. Complex fluids such as foams display an intermediate response to shear with nontrivial frequency-dependent shear moduli. In this manuscript, we conduct coordinated experiments and numerical simulations of model foams subjected to boundary-driven oscillatory, planar shear. Our studies are performed on bubble rafts (experiments) and the bubble model (simulations) in 2D. We focus on the low-amplitude flow regime in which T1 bubble rearrangement events do not occur, yet the system transitions from solid- to liquid-like behavior as the driving frequency is increased. In both simulations and experiments, we observe two distinct flow regimes. At low frequencies $ω$, the velocity profile of the bubbles increases linearly with distance from the stationary wall, and there is a nonzero total phase shift between the moving boundary and interior bubbles. In this frequency regime, the total phase shift scales as a power-law $Δ\sim ω^n$ with $n \approx 3$. In contrast, for frequencies above a crossover frequency $ω> ω_{p}$, the total phase shift $Δ$ scales linearly with the driving frequency. At even higher frequencies above a characteristic frequency $ω_{nl} > ω_{p}$, the velocity profile changes from linear to nonlinear. We fully characterize this transition from solid- to liquid-like flow behavior in both the simulations and experiments, and find qualitative and quantitative agreement for the characteristic frequencies.

cond-mat.soft↗