arXiv subjects

Shivam Tripathi

Publications and source records attributed to Shivam Tripathi.

4 recordsLinked to original sources

arXiv2026

Principia: Relational Physics Tests for Video Models

Evaluating physical reasoning in video models is difficult because absolute motion measurements depend on frame rate, object scale, and camera calibration, all of which are often ambiguous or unavailable in generated video. We propose a different approach. When two objects in the same scene obey the same physical law, their motions must satisfy predictable relationships, and these relationships hold independent of calibration. We introduce Principia, a benchmark that evaluates Newtonian physics through relational consistency between paired objects. Principia spans eight phenomena - gravity, restitution, friction, rotational inertia, projectile motion, momentum, pendulum, and mass-spring oscillation - across translational, rotational, collisional, and oscillatory dynamics, using real-world scenes recorded under controlled protocols. We also introduce a calibration-independent consistency score that quantifies physical violation directly in image space. Across thousands of generations from six state-of-the-art video generators, no model exceeds 0.42 on Principia despite all scoring around 0.8 on VBench. Vision-language models are evaluated on their ability to detect relational physics violations, with the best model achieving only 67% accuracy and most performing near chance level.

cs.CV↗

arXiv2026

Martensitic Transformation in Crystal-Amorphous Superlattices of NiTi Shape Memory Alloy

Shape memory alloys (SMAs) exhibit unique thermo-mechanical properties arising from reversible martensitic transformation. Tailoring these properties through coherent second phases is effective but is often limited by the difficulty of identifying thermodynamically compatible phases. Crystal-amorphous superlattices (CAS), in which the second phase is derived from the base material itself, provide an attractive alternative. However, the influence of partial amorphization on martensitic transformation and thermo-mechanical behavior remains largely unexplored. Here, large-scale molecular dynamics simulations are used to investigate temperature- and stress-induced martensitic transformations in CAS-NiTi with different crystalline phase fractions. Partial amorphization fundamentally modifies the transformation pathway, introducing an initial continuous (second-order-like) transformation before the conventional first-order martensitic transformation. The amorphous phase also enhances transformation reversibility, reducing the thermal hysteresis from 275 K in fully crystalline NiTi to 95-110 K in CAS-NiTi. Simultaneously, the elastic modulus and the critical stress for stress-induced martensitic transformation increase by approximately 60-90 percent, depending on the crystalline fraction. These improvements originate from heterogeneous nucleation at crystal-amorphous interfaces, retained austenite that promotes the reverse transformation, and mechanical constraint imposed by the amorphous phase. These findings establish crystal-amorphous superlattices as a promising microstructural design strategy for tailoring the thermo-mechanical performance of shape memory alloys.

cond-mat.mtrl-sci↗

arXiv2026

From Data-Driven Models to Physical Insight: Vibrational Entropy Governed by Atomic Volume

Vibrational entropy plays a central role in determining phase stability and temperature dependent behavior in materials, yet its calculation from first-principles phonon methods remains computationally demanding. In this work, we combine data-driven modeling with physically motivated analysis to develop an efficient and interpretable framework for predicting vibrational entropy. Using a dataset derived from PhononDB, a feedforward neural network trained on Materials Project and composition based descriptors achieves high predictive accuracy, while SHAP analysis identifies atomic volume as the dominant factor governing vibrational entropy. Guided by this insight, simplified analytical models are constructed, revealing a logarithmic dependence of vibrational entropy on atomic volume consistent with lattice dynamical considerations. A logarithmic linear model is shown to provide an accurate and physically interpretable description across the full range of materials. To extend the analysis to finite temperatures, a temperature dependent formulation is introduced that incorporates T3 scaling at low temperatures and logarithmic dependence at higher temperatures, consistent with Debye and Einstein type behavior. This unified model captures both structural and thermal contributions to vibrational entropy with good accuracy. Overall, the proposed framework demonstrates that vibrational entropy can be predicted using simple, physically meaningful relationships, offering a computationally efficient alternative to full phonon calculations and enabling entropy informed materials screening.

cond-mat.mtrl-sci↗

arXiv2025

Objects in Generated Videos Are Slower Than They Appear: Models Suffer Sub-Earth Gravity and Don't Know Galileo's Principle...for now

Video generators are increasingly evaluated as potential world models, which requires them to encode and understand physical laws. We investigate their representation of a fundamental law: gravity. Out-of-the-box video generators consistently generate objects falling at an effectively slower acceleration. However, these physical tests are often confounded by ambiguous metric scale. We first investigate if observed physical errors are artifacts of these ambiguities (e.g., incorrect frame rate assumptions). We find that even temporal rescaling cannot correct the high-variance gravity artifacts. To rigorously isolate the underlying physical representation from these confounds, we introduce a unit-free, two-object protocol that tests the timing ratio $t_1^2/t_2^2 = h_1/h_2$, a relationship independent of $g$, focal length, and scale. This relative test reveals violations of Galileo's equivalence principle. We then demonstrate that this physical gap can be partially mitigated with targeted specialization. A lightweight low-rank adaptor fine-tuned on only 100 single-ball clips raises $g_{\mathrm{eff}}$ from $1.81\,\mathrm{m/s^2}$ to $6.43\,\mathrm{m/s^2}$ (reaching $65\%$ of terrestrial gravity). This specialist adaptor also generalizes zero-shot to two-ball drops and inclined planes, offering initial evidence that specific physical laws can be corrected with minimal data.

cs.CV↗