Large Language Models (LLMs), through their contextualized representations, have been empirically proven to encapsulate syntactic, semantic, word sense, and common-sense knowledge. However, there has been limited exploration of their physical reasoning abilities, specifically concerning the crucial attributes for comprehending everyday objects. To address this gap, we introduce NEWTON, a repository and benchmark for evaluating the physics reasoning skills of LLMs. Further, to enable domain-specific adaptation of this benchmark, we present a pipeline to enable researchers to generate a variant of this benchmark that has been customized to the objects and attributes relevant for their application. The NEWTON repository comprises a collection of 2800 object-attribute pairs, providing the foundation for generating infinite-scale assessment templates. The NEWTON benchmark consists of 160K QA questions, curated using the NEWTON repository to investigate the physical reasoning capabilities of several mainstream language models across foundational, explicit, and implicit reasoning tasks. Through extensive empirical analysis, our results highlight the capabilities of LLMs for physical reasoning. We find that LLMs like GPT-4 demonstrate strong reasoning capabilities in scenario-based tasks but exhibit less consistency in object-attribute reasoning compared to humans (50\% vs. 84\%). Furthermore, the NEWTON platform demonstrates its potential for evaluating and enhancing language models, paving the way for their integration into physically grounded settings, such as robotic manipulation. Project site: https://newtonreasoning.github.io

NEWTON: Are Large Language Models Capable of Physical Reasoning?

Relational databases play an important role in business, science, and more. However, many users cannot fully unleash the analytical power of relational databases, because they are not familiar with database languages such as SQL. Many techniques have been proposed to automatically generate SQL from natural language, but they suffer from two issues: (1) they still make many mistakes, particularly for complex queries, and (2) they do not provide a flexible way for non-expert users to validate and refine incorrect queries. To address these issues, we introduce a new interaction mechanism that allows users to directly edit a step-by-step explanation of a query to fix errors. Our experiments on multiple datasets, as well as a user study with 24 participants, demonstrate that our approach can achieve better performance than multiple SOTA approaches. Our code and datasets are available at https://github.com/magic-YuanTian/STEPS.

Interactive Text-to-SQL Generation via Editable Step-by-Step Explanations | VIDEO

The task of table summarization involves generating text that both succinctly and accurately represents the table or a specific set of highlighted cells within a table. While significant progress has been made in table to text generation techniques, models still mostly generate descriptive summaries, which reiterates the information contained within the table in sentences. Through analysis of popular table to text benchmarks (ToTTo (Parikh et al., 2020 and InfoTabs (Gupta et al., 2020) we observe that in order to generate the ideal summary, multiple types of reasoning is needed coupled with access to knowledge beyond the scope of the table. To address this gap, we propose ReTAG, a table and reasoning aware model that uses vector-quantization to infuse different types of analytical reasoning into the output. ReTAG achieves 2.2%, 2.9% improvement on the PARENT metric in the relevant slice of ToTTo and InfoTabs for the table to text generation task over state of the art baselines. Through human evaluation, we observe that output from ReTAG is upto 12% more faithful and analytical compared to a strong table-aware model. To the best of our knowledge, ReTAG is the first model that can controllably use multiple reasoning methods within a structure-aware sequence to sequence model to surpass state of the art performance in multiple table to text tasks. We extend (and open source 35.6K analytical, 55.9k descriptive instances) the ToTTo, InfoTabs datasets with the reasoning categories used in each reference sentences.

ReTAG: Reasoning Aware Table to Analytic Text Generation

Although great progress has been made for Machine Reading Comprehension (MRC) in English, scaling out to a large number of languages remains a huge challenge due to the lack of large amounts of annotated training data in non-English languages. To address this challenge, some recent efforts of cross-lingual MRC employ machine translation to transfer knowledge from English to other languages, through either explicit alignment or implicit attention. For effective knowledge transition, it is beneficial to leverage both semantic and syntactic information. However, the existing methods fail to explicitly incorporate syntax information in model learning. Consequently, the models are not robust to errors in alignment and noises in attention. In this work, we propose a novel approach, named GraFusionMRC, which jointly models the cross-lingual alignment information and the mono-lingual syntax information using a graph. We develop a series of algorithms, including graph construction, learning, and pre-training. The experiments on two benchmark datasets for cross-lingual MRC show that our approach outperforms all strong baselines, which verifies the effectiveness of syntax information for cross-lingual MRC.

A Graph Fusion Approach for Cross-Lingual Machine Reading Comprehension

Although multilingual pretrained models (mPLMs) enabled support of various natural language processing in diverse languages, its limited coverage of 100+ languages lets 6500+ languages remain ‘unseen’. One common approach for an unseen language is specializing the model for it as target, by performing additional masked language modeling (MLM) with the target language corpus. However, we argue that, due to the discrepancy from multilingual MLM pretraining, a naive specialization as such can be suboptimal. Specifically, we pose three discrepancies to overcome. Script and linguistic discrepancy of the target language from the related seen languages, hinder a positive transfer, for which we propose to maximize representation similarity, unlike existing approaches maximizing overlaps. In addition, label space for MLM prediction can vary across languages, for which we propose to reinitialize top layers for a more effective adaptation. Experiments over four different language families and three tasks shows that our method improves the task performance of unseen languages with statistical significance, while previous approach fails to.

Script, Language, and Labels: Overcoming Three Discrepancies for Low-Resource Language Specialization

Ferrite-based junction circulators [1,2] are microwave devices made of at least three ports that have been widely used in emission/reception applications ensuring an
isolation between the incoming and outgoing signals. They are especially useful in single antenna applications where signals are being transmitted and received simultaneously.
Conventional circulators are based on a ferrite disk, magnetized to saturation, placed under the central conductor ensuring electromagnetic waves to propagate in a non-reciprocal manner.
The external magnetic field allowing this saturation is provided by a permanent magnet positioned right above of the central conductor.
Due to the size, weight, cost of these systems and the fact that high frequency applications are recently the main concern, the miniaturization of the circulator is inevitable. This can be
achieved using self-biased hexagonal ferrite materials [3,4] which eliminates the use of a permanent magnet. These self-polarized materials, suitable for high frequency applications, are
pellets made out of hexaferrite powder pressed and sintered. During pressing, the orientation of the particles should be achieved in the remanence state using an external magnetic field. In
order to avoid the application of a magnetic field during pressing, and thus to minimize the cost of manufacturing circulators, we propose to design and manufacture circulators based on
hexaferrite particles in the form of platelets. It has been shown that this platelet shape, as shown in Fig. 1, induces a self-alignment of the crystals with the particles orienting themselves
spontaneously, therefore dropping the application of an external magnetic field in any way. This material will be then characterized using a coaxial cell capable of determining its dielectric
and magnetic properties such as the permittivity and the permeability [5]. The values obtained will then be used in the simulation of a coplanar circulator [6,7], that operates in the 5G band
especially around 28 GHz, using HFSS simulator. 

**References ** 
1. H. Bosma, IEEE Transactions on Microwave Theory and Techniques, Vol 12(1), p. 61-72 (1964). 
2. Y. S. Wu and F. J. Rosenbaum. IEEE Transactions on Microwave Theory and Techniques Vol 22(10), p. 849-856 (1974). 
3. N. Zeina, H. How, C. Vittoria, In 1992 IEEE International Magnetics Conference (INTERMAG), pp. 458-458 (1992). 
4. N. Noutehou, C. Patris, D. Névo, Circulateur planaire ultra-compact en bande Q." In JNM (2019). 
5. F. Costa, M. Borgese, M. Degiorgi, In Electronics, Vol 6(4), p. 95 (2017). 
6. O. Zahwe,H. Harb and H. Nasrallah, Int. Arab J. Inf. Technol., Vol 16(3A), p.600-608 (2019). 
7. R. El Bouslemti,F. Salah-Belkhodja and Y. H. A. Fekhar, Revue Méditerranéenne des Télécommunications, Vol 4(1) (2014). 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/a9382678fd6176ba6e400762d0ad1e12.png)

Development of an Auto Oriented Circulator Based on Hexaferrite Materials in the 5G Frequency Band

One of the most important parts of a fluxgate sensor is the magnetic core. This work is dedicated to the study of fluxgate sensors based on the nanocrystalline wire as a core
and comparing them to the performance of amorphous wire with chemical composition (Co94Fe6)75Si15B10 [1]. Both wires were manufactured by the Institute of Technical Physics, Iasi,
Romania. Compared to the previous studies on nanocrystalline materials, we demonstrate that proper processing of nanocrystalline alloys can bring sensors with properties comparable to
the traditionally used amorphous alloys, but with better stability at elevated temperatures.
The amorphous Fe73,5Si13,5B9Cu1Nb3 master alloy wires 0.5m in length, and 0.12 mm in diameter were annealed by DC pulse to achieve the nanocrystalline structure. Contrary to other
researchers we found problems with surface oxidation in air, so our annealing was performed in Argon. During the annealing and cooling, we applied mechanical stress to achieve
transverse anisotropy. The character of changing Hc varies depending on tension as shown in Fig.1. Higher tension increases anisotropy and reduces the longitudinal permeability (Fig. 2).
Best results were shown by wires annealed under tension 9 g and 13 g with the current in the range 485 mA up to 510 mA. In these cases, coercivity was in the range of 7 A/m up to 17
A/m at 1 kHz. Annealed wires were used as a core for fluxgate sensors of the Förster and Vacquier types and with ring core.
As an example, Förster type sensor with amorphous (Hc = 20 A/m) wire as a core had max. sensitivity of 0.88 V/mT and min. the noise of 164pT/√Hz at 1 Hz.
By using nanocrystalline wire with Hc = 17A/m we achieved lower sensitivity (0.7 V/mT) but lower noise (111 pT/√Hz). The nanocrystalline core with lower Hc (9 A/m) had higher
sensitivity (1.1 V/mT) but higher noise (204 pT/√Hz) compared to the amorphous core. And the core with Hc = 13A/m had almost the same results as the amorphous wire.
This study was supported by the Grant Agency of the Czech Republic within the Nanofluxgate project (GACR GA20-27150S). 

**References:** 

[1] P. Ripka, D. Hrakova, “Multiwire Parallel Fluxgate Sensors”, IEEE Transactions on Magnetics. 2022, 58(2), 1-5. ISSN 0018-9464. DOI 10.1109/TMAG.2021.3093017 
[2] Y.-F. Li, M. Vazquez, D.-X. Chen, “Circular magnetization process of nanocrystalline wires as deduced from impedance measurements”, Journal of Applied Physics 97, 124311 (2005) 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/dc0c8aa0034b27a0fcea032449a5d896.png)

Creep induced Magnetic Anisotropy of Nanocrystalline Wire as a Core of Fluxgate Sensor

Susceptibility measurement of magnetic nanoparticles (MNPs) in radio frequency (RF) has been widely studied. However, most of the studies only examined frequencies
up to 20 GHz [1]. In addtion, applying two axial DC fields to increase resonance in specific frequency range and magnetic susceptibility has hardly been done before, since applying a
single static DC field is more common [2-3]. The MNPs can be used for magnetic hyperthermia and biological substances detection [4-6]. In this study, we proposed a novel method for
measuring magnetic susceptibility by a broadband microstrip line type probe up to 67 GHz [7]. Fig. 1 shows the basic principle of the experimental setup. The microtube with MNPs
(SVM-025-5H, 200 nmf, about 300 ml) was put on the microstrip line type probe, which is connected with a network analyzer (N5227A) to generate a magnetic field in a range of 10 MHz
to 67 GHz. The MNPs were put between two electromagnets connected with power supplies to generate a changeable DC field up to 1000 Oe and another orthogonal about 35 Oe DC field.
When the two orthogonal DC fields are around equal, the magnetic susceptibility of MNPs should maximize because of the magnetic switching theory [8], which results in enhancing the
signal to noise ratio. By applying the bias magnetic field, we measured the transmission coefficient, S21, of the microstrip line type probe and obtained magnetic susceptibility [7]. Fig. 2
shows the magnetic susceptibility when the DC field ranges from 0 Oe to 1000 Oe. In the DC field around 0 Oe, the magnetic susceptibility was very low because the MNPs were relaxed.
However, in the DC field of 25 Oe-75 Oe, we observed clear magnetic resonances in 20 - 40 GHz. The resonance frequency increased as the DC field increased. In this case, both fixed and
changeable DC fields are around 35 Oe, which means the magnetic susceptibility is biggest at this condition and the orthogonal DC field enhance the signal to noise ratio to detect magnetic
nanoparticle. The clear magnetic resonance of MNPs has hardly been observed in this broad bandwidth up to 67 GHz before. 


**References:** 

[1] B. K. Kuanr, V. Veerakumar and K. Lingam, Journal of Applied Physics, Vol. 105 (2009). 
[2] M. Jadav, S. P. Bhatnagar, IEEE Transactions on Magnetics, Vol. 56, p. 1-8 (2020). 
[3] P. C. Fannin, I. Malaescu and C. N. Marin, The European Physical Journal E, Vol. 27, p. 145-148 (2008). 
[4] A. Shikano, L. Tonthat, Transaction of the Magnetics Society of Japan Special Issues, Vol. 6, p. 100-104 (2022). 
[5] S. Yabukami, T. Murayama and S. Takahashi, IEEE Transactions on Magnetics, Vol. 58, 6100305 (2022, in press). 
[6] T. Yoneyama, A. Kuwahata and T. Murayama, IEEE Transactions on Magnetics, Vol. 58 (2022, in press). 
[7] S. Yabukami, C. Iwasaki and K. Nozawa, IEEE Transactions on Magnetics, Vol. 58, No. 2, p. 1-5 (2022) 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/715b881ce9b3e0d6f065037d7f9ab76f.png)

Method for Measuring Magnetic Susceptibility of Magnetic Nanoparticles up to Millimeter Wave Frequency Range

The usage of antibody selective biding for the detection of biomolecules provides a fast and low-cost testing platform. The most common type of antibody selective biding
bio-detection tags the bio-molecule of interest with a visual marker. An example of this type of bio-detection is the lateral flow immunoassays like the pregnancy or rapid COVID tests [1].
These visual nanotags are a valid solution for binary detection in high concentrated samples. To have a quantifiable and high sensitivity bio-sensing we use magnetic nanoparticles. The
measurements of this type of magnetic nanotags require expensive laboratory equipment. A novel method using the self-resonant frequency (SRF) of inductors allows high sensitivity with
low-cost and portable systems.
Using the equivalent impedance calculation [2] and the Lichtenecker’s mixing formulas [3] we extracted an analytic simplification to describe the relation between masses of nanomaterial
in the sensing area to variations of the SRF of the inductor. The model correlates variations of the refraction index of the nanotag making the technique able to quantify not only magnetic
nanotags but other types of nanomaterials like gold or latex.
The mathematical model and its simplifications are tested against electromagnetic simulations and experimental measurements. These verifications demonstrate the validity of the model,
the value of magnetic nanoparticles as nanotags for precise quantification, and the advantages of the SRF measuring method for low-cost, high-sensitivity systems. 

**References:** 

1] Moyano, A. et al., 2020. Magnetic Lateral Flow Immunoassays. Diagnostics. 
[2] Qin Yu. et al., 2002. RF Equivalent Circuit Modeling of Ferrite-Core Inductors and Characterization of Core Materials. 
[3] A.V. Goncharenko. et al., 1999. Lichtenecker’s equation: applicability and limitations. 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/f53b5fd75dba2f666b231ac46824d773.png)

Self resonant frequency of inductors for quantification of magnetic nanoparticles

Permanents magnets, or hard ferromagnets, provide a “free” source of magnetic field, with no need of external power. Considering the integration with microsystems,
permanent micromagnets are critical components for the fabrication of microscale motors, generators, switches, pumps, acoustic speakers and energy harvester. Even though the
understanding of magnetism and magnetic materials has led to advancements in their magnetic properties [1], the application of permanent magnets in microsystems is still limited. The
reasons for that are: i) the methods for magnet fabrication at the macroscale (casting and powder processing) are fundamentally different to thin-film microfabrication approaches (physical
vapor deposition, chemical vapor deposition, electrochemical deposition); ii) even if microfabrication approaches are exploited, difficulties remain in depositing thick magnetic films (> 1
μm) and issues due to process compatibility [2].
In this work, we present a process to integrate MEMS (MicroElectroMechanical Systems) and thick permanent magnets. We have designed test structures (Fig. 1) and optimized a process
to fabricate the devices. The optimizations concern: i) deposition of a micron-thick layer of a magnetic material; ii) thermal treatment to enhance the magnetic characteristics of the
material; iii) patterning technique employed to obtain micromagnets; iv) introduction of a passivation layer to protect the material from the subsequent process steps. An image of the final
fabricated device is shown in Fig.2. Examples of application of said process to devices for energy harvesting and magnetic sensing will be discussed. 

**References:** 

M. Sagawa, S. Fujimura, N. Togawa, Journal of Applied Physics, Vol. 55, p. 2083 (1984) 
D. P. Arnold, N. Wang, Journal of microelectromechanical systems, Vol. 18, p.1255 (2009) 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/2e8453b58011aaaa96750f6eb42a03f7.png)

Integration of hard magnetic layer in MEMS process

Tunable perpendicular magnetic anisotropy (PMA) in thin film magnetic devices provides an isotropic in-plane permeability, necessary for integrated microwave and mm-wave devices and an opportunity to break inversion symmetry that allows implementations of non-reciprocal devices like circulators and isolators. In ferrimagnetic alloys like GdxCo1-x, this is usually achieved by changing the composition such that the magnetic properties lie on either side of the composition-dependent magnetic compensation at room temperature, as shown in Fig. 1. Different compositions achieved in this work by changing the sputtering power of the Gd target show that magnetic compensation window lies between x = 28-32, and the effective anisotropy diverges about the compensation point. 

One less explored pathway is the incorporation of oxygen, which until now has been ruled out as a detriment to the ferrimagnetic properties [1]. Here, we show that preferential oxygenation of GdxCo1-x through a reactive sputtering process, such that the cation (Gd and Co) to oxygen anion ratio is 1:1, increases the effective anisotropy energy density by two orders of magnitude compared to the corresponding alloys without oxygen. X-ray photoelectron spectroscopy reveals that strong PMA with an optimal oxygen flow results in a Co-deficient stoichiometry of Gd21Co28O51, which suggests that strong ferrimagnetic order likely arises from a superexchange-like coupling between Gd and Co via the non-magnetic O [2]. This can produce a spontaneous magnetization in the out-of-plane direction due to weaker exchange interactions within the alloy in a heavy-metal/ferrimagnet heterostructure for PMA. Even greater anisotropy fields on the order of 11 kOe are achieved in a super-lattice with 10 repetitions of the GdxCo1-x heterostructure unit-cell, which corresponds to an operating frequency of 30 GHz. Therefore, not only does oxygenation increase PMA and provide tunable anisotropy to amorphous Gd-Co films, but the oxygen incorporation itself also increases the resistivity of the film, making it even more suitable for that frequency range. 
**References:** [1] K. Ueda, A. J. Tan, and G. S. D. Beach, AIP Advances, vol. 8, no. 12, p. 125204, Dec. 2018. 
[2] R. J. Gambino and J. J. Cuomo, Journal of Vacuum Science and Technology, vol. 15, no. 2, pp. 296–301, Mar. 1978. 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/0641d84a745d89c9edb1bbc17e16653a.png) 
Fig. 1. Saturation Magnetization and Effective Anisotropy Energy Density for varying GdxCo1-x compositions for x = 18-39.

Role of Composition and Oxygen in Ferrimagnetic GdxCo1 x Alloys for Microwave/mm

High-performance magneto-dielectrics with high permeability and permittivity, low magnetic and dielectric losses, and high cutoff frequency are promising candidate materials for next-generation communication applications, owing to their significant advantages for antenna miniaturization and bandwidth enhancement. Due to limited resonant frequencies and large natural resonance linewidths, conventional microwave ferrites and their composites, which are usually used over the frequency range from several hundred MHz to several GHz, do not meet the requirements for 5G communications.1-3 A unique type of hexaferrites named Mg-Zn 18H hexaferrite, whose crystal structure can be interpreted as the structure of Y-type hexaferrites with intercalation of a three-layered hexagonal barium titanate block into the middle of the T-block as shown in Fig. 1a, has been synthesized, and its microwave properties have been comprehensively investigated for the first time.4 The Mg-Zn 18H hexaferrite shows an extremely low damping coefficient of 0.1-0.2, indicating concentrated magnetic loss within a narrow frequency band and a narrow ferrimagnetic resonance linewidth of 486-660 Oe. Owing to this remarkably low damping coefficient, the Mg-Zn 18H hexaferrite possesses an excellent magnetic loss tangent as low as 0.06 and a small dielectric loss tangent less than 0.006 over 2-4 GHz, representing superior magneto-dielectric performance among the reported microwave ferrites for the S band applications as shown in Fig. 1b. Moreover, excellent thermal stability for the damping coefficient is observed to be 0.0004 K-1 over the temperature variation of most practical applications. When employed as the substrate of a 3.6-GHz patch antenna, the Mg-Zn 18H hexaferrite shows a miniaturization factor of ~5 and significant bandwidth improvement of ~50-110% over conventional dielectrics. These results imply the great technological potential and commercial value of the 18H hexaferrites for 5G communications. 
**References:** 1V. G. Harris, IEEE Trans. Mag. 48, 1075 (2012). 
2V. G. Harris, A. Geiler, Y. Chen, S. Yoon, M. Wu, A. Yang, Z. Chen, P. He, P. V. Parimi, X. Zuo, C. E. Patton, M. Abe, O. Acher, and C. Vittoria, J. Magn. Magn. Mater. 321, 2035 (2009). 
3Q. Li, Y. Chen, C. Yu, K. Qian, V. G. Harris, J. Am. Ceram. Soc. 103, 5076 (2020). 
4Q. Li, Y. Chen, C. Yu, L, Young, J. Spector, and V. G. Harris, Acta Mater. 231, 117854 (2022). 

![](https://s3.eu-west-1.amazonaws.com/underline.prod/uploads/markdown_image/1/image/b2b57c67782800adf088a5c437a9d3be.png) 
(a) Schematic illustration of the Mg 18H hexaferrite unit cell with the stacking subunit blocks denoted. (b) Performance of magneto-dielectric materials for microwave applications. The figure of merit (FOM) are indicated by the dashed lines.

Underline digital video library

NEWTON: Are Large Language Models Capable of Physical Reasoning?

Interactive Text-to-SQL Generation via Editable Step-by-Step Explanations | VIDEO

ReTAG: Reasoning Aware Table to Analytic Text Generation

A Graph Fusion Approach for Cross-Lingual Machine Reading Comprehension

Script, Language, and Labels: Overcoming Three Discrepancies for Low-Resource Language Specialization

Development of an Auto Oriented Circulator Based on Hexaferrite Materials in the 5G Frequency Band

Creep induced Magnetic Anisotropy of Nanocrystalline Wire as a Core of Fluxgate Sensor

Method for Measuring Magnetic Susceptibility of Magnetic Nanoparticles up to Millimeter Wave Frequency Range

Self resonant frequency of inductors for quantification of magnetic nanoparticles

Integration of hard magnetic layer in MEMS process

Role of Composition and Oxygen in Ferrimagnetic GdxCo1 x Alloys for Microwave/mm

Emerging Magneto Dielectric Materials for 5G Antenna Miniaturization: 18H Hexaferrites

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES