Showing 1–15 of 15 results for author: Soria, A

Granite-Function Calling Model: Introducing Function Calling Abilities via Multi-task Learning of Granular Tasks

Authors: Ibrahim Abdelaziz, Kinjal Basu, Mayank Agarwal, Sadhana Kumaravel, Matthew Stallone, Rameswar Panda, Yara Rizk, GP Bhargav, Maxwell Crouse, Chulaka Gunasekara, Shajith Ikbal, Sachin Joshi, Hima Karanam, Vineet Kumar, Asim Munawar, Sumit Neelam, Dinesh Raghu, Udit Sharma, Adriana Meza Soria, Dheeraj Sreedhar, Praveen Venkateswaran, Merve Unuvar, David Cox, Salim Roukos, Luis Lastras , et al. (1 additional authors not shown)

Abstract: Large language models (LLMs) have recently shown tremendous promise in serving as the backbone to agentic systems, as demonstrated by their performance in multi-faceted, challenging benchmarks like SWE-Bench and Agent-Bench. However, to realize the true potential of LLMs as autonomous agents, they must learn to identify, call, and interact with external tools and application program interfaces (AP… ▽ More Large language models (LLMs) have recently shown tremendous promise in serving as the backbone to agentic systems, as demonstrated by their performance in multi-faceted, challenging benchmarks like SWE-Bench and Agent-Bench. However, to realize the true potential of LLMs as autonomous agents, they must learn to identify, call, and interact with external tools and application program interfaces (APIs) to complete complex tasks. These tasks together are termed function calling. Endowing LLMs with function calling abilities leads to a myriad of advantages, such as access to current and domain-specific information in databases and knowledge sources, and the ability to outsource tasks that can be reliably performed by tools, e.g., a Python interpreter or calculator. While there has been significant progress in function calling with LLMs, there is still a dearth of open models that perform on par with proprietary LLMs like GPT, Claude, and Gemini. Therefore, in this work, we introduce the GRANITE-20B-FUNCTIONCALLING model under an Apache 2.0 license. The model is trained using a multi-task training approach on seven fundamental tasks encompassed in function calling, those being Nested Function Calling, Function Chaining, Parallel Functions, Function Name Detection, Parameter-Value Pair Detection, Next-Best Function, and Response Generation. We present a comprehensive evaluation on multiple out-of-domain datasets comparing GRANITE-20B-FUNCTIONCALLING to more than 15 other best proprietary and open models. GRANITE-20B-FUNCTIONCALLING provides the best performance among all open models on the Berkeley Function Calling Leaderboard and fourth overall. As a result of the diverse tasks and datasets used for training our model, we show that GRANITE-20B-FUNCTIONCALLING has better generalizability on multiple tasks in seven different evaluation datasets. △ Less

Submitted 27 June, 2024; originally announced July 2024.

Granite Code Models: A Family of Open Foundation Models for Code Intelligence

Authors: Mayank Mishra, Matt Stallone, Gaoyuan Zhang, Yikang Shen, Aditya Prasad, Adriana Meza Soria, Michele Merler, Parameswaran Selvam, Saptha Surendran, Shivdeep Singh, Manish Sethi, Xuan-Hong Dang, Pengyuan Li, Kun-Lung Wu, Syed Zawad, Andrew Coleman, Matthew White, Mark Lewis, Raju Pavuluri, Yan Koyfman, Boris Lublinsky, Maximilien de Bayser, Ibrahim Abdelaziz, Kinjal Basu, Mayank Agarwal , et al. (21 additional authors not shown)

Abstract: Large Language Models (LLMs) trained on code are revolutionizing the software development process. Increasingly, code LLMs are being integrated into software development environments to improve the productivity of human programmers, and LLM-based agents are beginning to show promise for handling complex tasks autonomously. Realizing the full potential of code LLMs requires a wide range of capabili… ▽ More Large Language Models (LLMs) trained on code are revolutionizing the software development process. Increasingly, code LLMs are being integrated into software development environments to improve the productivity of human programmers, and LLM-based agents are beginning to show promise for handling complex tasks autonomously. Realizing the full potential of code LLMs requires a wide range of capabilities, including code generation, fixing bugs, explaining and documenting code, maintaining repositories, and more. In this work, we introduce the Granite series of decoder-only code models for code generative tasks, trained with code written in 116 programming languages. The Granite Code models family consists of models ranging in size from 3 to 34 billion parameters, suitable for applications ranging from complex application modernization tasks to on-device memory-constrained use cases. Evaluation on a comprehensive set of tasks demonstrates that Granite Code models consistently reaches state-of-the-art performance among available open-source code LLMs. The Granite Code model family was optimized for enterprise software development workflows and performs well across a range of coding tasks (e.g. code generation, fixing and explanation), making it a versatile all around code model. We release all our Granite Code models under an Apache 2.0 license for both research and commercial use. △ Less

Submitted 7 May, 2024; originally announced May 2024.

Comments: Corresponding Authors: Rameswar Panda, Ruchir Puri; Equal Contributors: Mayank Mishra, Matt Stallone, Gaoyuan Zhang

API Pack: A Massive Multi-Programming Language Dataset for API Call Generation

Authors: Zhen Guo, Adriana Meza Soria, Wei Sun, Yikang Shen, Rameswar Panda

Abstract: We introduce API Pack, a massive multi-programming language dataset containing more than 1 million instruction-API call pairs to improve the API call generation capabilities of large language models. By fine-tuning CodeLlama-13B on 20,000 Python instances from API Pack, we enable it to outperform GPT-3.5 and GPT-4 in generating unseen API calls. Fine-tuning on API Pack also facilitates cross-progr… ▽ More We introduce API Pack, a massive multi-programming language dataset containing more than 1 million instruction-API call pairs to improve the API call generation capabilities of large language models. By fine-tuning CodeLlama-13B on 20,000 Python instances from API Pack, we enable it to outperform GPT-3.5 and GPT-4 in generating unseen API calls. Fine-tuning on API Pack also facilitates cross-programming language generalization by leveraging a large amount of data in one language and small amounts of data from other languages. Scaling the training data to 1 million instances further improves the model's ability to generalize to new APIs not used in training. To facilitate further research, we open-source the API Pack dataset, trained model, and associated source code at https://github.com/zguo0525/API-Pack. △ Less

Submitted 3 June, 2024; v1 submitted 14 February, 2024; originally announced February 2024.

New spinorial mass-quasilocal angular momentum inequality for initial data with marginally future trapped surface

Authors: Jarosław Kopiński, Alberto Soria, Juan A. Valiente Kroon

Abstract: We prove a new geometric inequality that relates the Arnowitt-Deser-Misner mass of initial data to a quasilocal angular momentum of a marginally future trapped surface inner boundary. The inequality is expressed in terms of a 1-spinor, which satisfies an intrinsic first-order Dirac-type equation. Furthermore, we show that if the initial data is axisymmetric, then the divergence-free vector used to… ▽ More We prove a new geometric inequality that relates the Arnowitt-Deser-Misner mass of initial data to a quasilocal angular momentum of a marginally future trapped surface inner boundary. The inequality is expressed in terms of a 1-spinor, which satisfies an intrinsic first-order Dirac-type equation. Furthermore, we show that if the initial data is axisymmetric, then the divergence-free vector used to define the quasilocal angular momentum cannot be a Killing field of the generic boundary. △ Less

Submitted 16 April, 2024; v1 submitted 30 October, 2023; originally announced October 2023.

Comments: 13 pages

Journal ref: J. Math. Phys. 65, 042501 (2024)

Ruled surfaces in 3-dimensional Riemannian manifolds

Authors: Marco Castrillón, María Eugenia Rosado, Alberto Soria

Abstract: In this work, ruled surfaces in 3-dimensional Riemannian manifolds are studied. We determine the expression for the extrinsic and sectional curvature of a parametrized ruled surface, where the former one is shown to be non-positive. We also quantify the set of ruling vector fields along a given base curve which allows to define a relevant reference frame that we refer to as Sannia frame. The funda… ▽ More In this work, ruled surfaces in 3-dimensional Riemannian manifolds are studied. We determine the expression for the extrinsic and sectional curvature of a parametrized ruled surface, where the former one is shown to be non-positive. We also quantify the set of ruling vector fields along a given base curve which allows to define a relevant reference frame that we refer to as Sannia frame. The fundamental theorem of existence and equivalence of Sannia-ruled surfaces in terms of a system of invariants is given. The second part of the article tackles the concept of the striction curve, which is proven to be the set of points where the so-called Jacobi evolution function vanishes on a ruled surface. This characterization of striction curves provides independent proof for their existence and uniqueness in space forms and disproves their existence or uniqueness in some other cases. △ Less

Submitted 18 December, 2023; v1 submitted 24 July, 2023; originally announced July 2023.

Comments: 22 pages

MSC Class: 53B25; 53B20; 53A55

Earthquake Occurrences in the Pacific Ring of Fire Exhibit a Collective Stochastic Memory for Magnitudes, Depths, and Relative Distances of Events

Authors: Pamela Jessica C. Roque, Renante R. Violanda, Christopher C. Bernido, Janneli Lea A. Soria

Abstract: Around 90% of the earthquakes in the world occur at the circum-Pacific belt referred to as the Pacific Ring of Fire exposing the countries in this region to high risk of earthquake hazards. We model fluctuations of the different seismic magnitudes, interevent distances, and seismic depths as a function of earthquake occurrence from the earthquake catalogs of Chile, Mexico, Japan, New Zealand, Phil… ▽ More Around 90% of the earthquakes in the world occur at the circum-Pacific belt referred to as the Pacific Ring of Fire exposing the countries in this region to high risk of earthquake hazards. We model fluctuations of the different seismic magnitudes, interevent distances, and seismic depths as a function of earthquake occurrence from the earthquake catalogs of Chile, Mexico, Japan, New Zealand, Philippines, and Southern California as a stochastic process with long-term memory. We show that the fluctuations of the three seismic quantities mentioned for all regions studied in this paper are governed by a single memory function that is described by a memory parameter μ and a decay parameter \b{eta}. The values of μ exhibit an underlying characteristic memory behavior of seismic activities common to all the countries considered, while the values of \b{eta} suggest a regional dependence which could be a manifestation of different seismic dynamics in various regions. This new perspective may provide a more versatile approach in studying the independent datasets that may be extracted from various earthquake catalogs. △ Less

Submitted 1 April, 2023; originally announced April 2023.

Comments: 11 Figures

Spacelike hypersurfaces in twisted product spacetimes with complete fiber and Calabi-Bernstein-type problems

Authors: Alberto Soria

Abstract: In this article spacelike hypersurfaces immersed in twisted product spacetimes $I\times_f F$ with complete fiber are studied. Several conditions ensuring global hyperbolicity are presented, as well as a relation that needs to hold on each spacelike hypersurface in $I\times_f F$ for it to be a simple warped product. When the fiber is assumed to be closed (compact and without boundary) and the ambie… ▽ More In this article spacelike hypersurfaces immersed in twisted product spacetimes $I\times_f F$ with complete fiber are studied. Several conditions ensuring global hyperbolicity are presented, as well as a relation that needs to hold on each spacelike hypersurface in $I\times_f F$ for it to be a simple warped product. When the fiber is assumed to be closed (compact and without boundary) and the ambient spacetime has a suitable expanding behaviour, non-existence results for constant mean curvature hypersurfaces are obtained. Under the same hypothesis, a characterization of compact maximal hypersurfaces and other for totally umbilic ones with a suitable restriction on their mean curvature are presented. The description of maximal hypersurfaces in twisted product spacetimes of the form $I\,{ }_{f}\!\!\times F$ with a one-dimensional Lorentzian fiber is also included. Finally, the mean curvature equation for a spacelike graph on the fiber is computed and as an application, some Calabi-Bernstein-type results are proven. We also include in an Appendix some known conformal geometry results describing the transformation of relevant tensors and operators under the action of a conformal map in a pseudo-Riemannian background. △ Less

Submitted 15 November, 2022; originally announced November 2022.

Comments: 23 pages

MSC Class: 53C50; 53C42; 53C80

Binding Group of Oligonucleotides on TiO 2 Surfaces: Phosphate Anions or Nucleobases?

Authors: Federico A. Soria, Cristiana Di Valentin

Abstract: Although the immobilization of oligonucleotides (nucleic acid) on mineral surfaces is at the basis of different biotechnological applications, an atomistic understanding of the interaction of the nucleic acid components with the titanium dioxide surfaces has not yet been achieved. Here, the adsorption of the phosphate anion, of the four DNA bases (adenine, guanine, thymine, and cytosine) and of so… ▽ More Although the immobilization of oligonucleotides (nucleic acid) on mineral surfaces is at the basis of different biotechnological applications, an atomistic understanding of the interaction of the nucleic acid components with the titanium dioxide surfaces has not yet been achieved. Here, the adsorption of the phosphate anion, of the four DNA bases (adenine, guanine, thymine, and cytosine) and of some entire nucleotides and dinucleotides on the TiO 2 anatase (101) surface is studied through dispersion-corrected hybrid density functional theory (DFT) calculations. Several adsorption configurations are identified for the separated entities (phosphate anion or base) and then considered when studying the adsorption of the entire nucleotides. The analysis shows that both the phosphate anion and each base may anchor the nucleotides to the surface in a collaborative and synergistic adsorption mode. The tendency is that the nucleotides containing the guanine base present the strongest adsorption while those made up with the thymine base have the lowest adsorption energies. Nucleotides based on adenine and cytosine have a similar intermediate behavior. Finally, we investigated the adsorption of competing water molecules to understand whether in the presence of the aqueous solvent, the nucleotides would remain bonded to the surface or desorb. △ Less

Submitted 10 November, 2021; v1 submitted 25 October, 2021; originally announced October 2021.

A Synchrotron as Accelerator of Science Development in Central America and the Caribbean

Authors: Galileo Violini, VÍctor M. Castaño, Juan Alfonso Fuentes Soria, Plácido Gómez Ramírez, Gregorio Medrano Asensio, Eduardo Posada, Carlos Rudamas

Abstract: Central America and the Caribbean (CAC) need science development efforts through ambitious projects that require strong regional collaboration. Inspiration can be drawn from initiatives in regions with similar problems. The bottleneck is the scarcity of public research centers and little or no research in private universities. An interesting proposal is the creation of a Dominican "Silicon Beach".… ▽ More Central America and the Caribbean (CAC) need science development efforts through ambitious projects that require strong regional collaboration. Inspiration can be drawn from initiatives in regions with similar problems. The bottleneck is the scarcity of public research centers and little or no research in private universities. An interesting proposal is the creation of a Dominican "Silicon Beach". The "Central American Science and Technology Fund" should focus on objectives capable of attracting the attention of the non-academic sector, first and foremost policy makers, but also civil society in general. The successful experience of SESAME (" Synchrotron Light for Experimental Science and Applications in Middle East ") offers an interesting basis for reflection, as it allows scientific research and short-term practical and social applications. Only two of the more than 60 existing synchrotrons are in Latin America, both in Brazil. Together with other similar projects in the South, such as the African Light Source (AFLS), and with the support of SESAME, LNLS and other synchrotrons in the South, it could lead to interesting South-South cooperation, which could be supported by the European Union or the NSF.As David Gross reminded, Science drives Technology, Technology drives Innovation, and this ends up in the welfare of society. A regional synchrotron may be the way to make this a reality in the Great Caribbean Region, as a first historical example of a large regional facility there. △ Less

Submitted 3 December, 2023; v1 submitted 24 September, 2021; originally announced September 2021.

On the Penrose inequality along null hypersurfaces

Authors: Marc Mars, Alberto Soria

Abstract: The null Penrose inequality, i.e. the Penrose inequality in terms of the Bondi energy, is studied by introducing a funtional on surfaces and studying its properties along a null hypersurface $Ω$ extending to past null infinity. We prove a general Penrose-type inequality which involves the limit at infinity of the Hawking energy along a specific class of geodesic foliations called Geodesic Asymptot… ▽ More The null Penrose inequality, i.e. the Penrose inequality in terms of the Bondi energy, is studied by introducing a funtional on surfaces and studying its properties along a null hypersurface $Ω$ extending to past null infinity. We prove a general Penrose-type inequality which involves the limit at infinity of the Hawking energy along a specific class of geodesic foliations called Geodesic Asymptotic Bondi (GAB), which are shown to always exist. Whenever, this foliation approaches large spheres, this inequality becomes the null Penrose inequality and we recover the results of Ludvigsen-Vickers and Bergqvist. By exploiting further properties of the functional along general geodesic foliations, we introduce an approach to the null Penrose inequality called Renormalized Area Method and find a set of two conditions which implies the validity of the null Penrose inequality. One of the conditions involves a limit at infinity and the other a condition on the spacetime curvature along the flow. We investigate their range of applicability in two particular but interesting cases, namely the shear-free and vacuum case, where the null Penrose inequality is known to hold from the results by Sauter, and the case of null shells propagating in the Minkowski spacetime. Finally, a general inequality bounding the area of the quasi-local black hole in terms of an asymptotic quantity intrinsic of $Ω$ is derived. △ Less

Submitted 19 November, 2015; originally announced November 2015.

Comments: 32 pages, 0 figures

The asymptotic behaviour of the Hawking energy along null asymptotically flat hypersurfaces

Authors: Marc Mars, Alberto Soria

Abstract: In this work we obtain the limit of the Hawking energy of a large class of foliations along general null hypersurfaces $Ω$ satisfying a weak notion of asymptotic flatness. The foliations are not required to be either geodesic or approaching large spheres at infinity. The limit is obtained in terms of a reference background geodesic foliation approaching large spheres and a positive function, const… ▽ More In this work we obtain the limit of the Hawking energy of a large class of foliations along general null hypersurfaces $Ω$ satisfying a weak notion of asymptotic flatness. The foliations are not required to be either geodesic or approaching large spheres at infinity. The limit is obtained in terms of a reference background geodesic foliation approaching large spheres and a positive function, constant along the null generators on $Ω$, which describes the relation between the two foliations at infinity. The integrand in the limit expression has interesting covariance and invariance properties with respect to change of background foliation. The standard result that the Hawking energy tends to the Bondi energy under suitable circumstances is recovered in this framework. △ Less

Submitted 4 June, 2015; originally announced June 2015.

Comments: 29 pages, no figures

MSC Class: 83C30; 83C40

Faster-is-slower effect in esca** ants revisited: Ants do not behave like humans

Authors: Daniel R. Parisi, Sabrina A Soria, Roxana Josens

Abstract: In this work we studied the trajectories, velocities and densities of ants when egressing under controlled levels of stress produced by a chemical repellent at different concentrations. We found that, unlike other animals esca** under life-and-death conditions and pedestrian simulations, ants do not produce a higher density zone near the exit door. Instead, ants are uniformly distributed over th… ▽ More In this work we studied the trajectories, velocities and densities of ants when egressing under controlled levels of stress produced by a chemical repellent at different concentrations. We found that, unlike other animals esca** under life-and-death conditions and pedestrian simulations, ants do not produce a higher density zone near the exit door. Instead, ants are uniformly distributed over the available space allowing for efficient evacuations. Consequently, the faster-is-slower effect observed in ants (Soria et al., 2012) is clearly of a different nature to that predicted by de social force model. In the case of ants, the minimum evacuation time is correlated with the lower probability of taking backward steps. Thus, as biological model ants have important differences that make their use inadvisable for the design of human facilities. △ Less

Submitted 20 October, 2014; originally announced October 2014.

Journal ref: Safety Science 72, 2015, 274-282

Geometry of normal graphs in Euclidean space and applications to the Penrose inequality in Minkowski

Authors: Marc Mars, Alberto Soria

Abstract: The Penrose inequality in Minkowski is a geometric inequality relating the total outer null expansion and the area of closed, connected and spacelike codimension-two surfaces S in the Minkowski spacetime, subject to an additional convexity assumption. In a recent paper, Brendle and Wang find a sufficient condition for the validity of this Penrose inequality in terms of the geometry of the orthogon… ▽ More The Penrose inequality in Minkowski is a geometric inequality relating the total outer null expansion and the area of closed, connected and spacelike codimension-two surfaces S in the Minkowski spacetime, subject to an additional convexity assumption. In a recent paper, Brendle and Wang find a sufficient condition for the validity of this Penrose inequality in terms of the geometry of the orthogonal projection of S onto a constant time hyperplane. In this work, we study the geometry of hypersurfaces in n-dimensional euclidean space which are normal graphs over other surfaces and relate the intrinsic and extrinsic geometry of the graph with that of the base hypersurface. These results are used to rewrite Brendle and Wang's condition explicitly in terms of the time height function of S over a hyperplane and the geometry of the projection of S along its past null cone onto this hyperplane. We also include, in an Appendix, a self-contained summary of known and new results on the geometry of projections along the Killing direction of codimension two-spacelike surfaces in a strictly static spacetime. △ Less

Submitted 19 July, 2013; originally announced July 2013.

Comments: 15 pages, 1 figure, Latex

On the Penrose inequality for dust null shells in the Minkowski spacetime of arbitrary dimension

Authors: Marc Mars, Alberto Soria

Abstract: A particular, yet relevant, particular case of the Penrose inequality involves null shells propagating in the Minkowski spacetime. Despite previous claims in the literature, the validity of this inequality remains open. In this paper we rewrite this inequality in terms of the geometry of the surface obtained by intersecting the past null cone of the original surface S with a constant time hyperpla… ▽ More A particular, yet relevant, particular case of the Penrose inequality involves null shells propagating in the Minkowski spacetime. Despite previous claims in the literature, the validity of this inequality remains open. In this paper we rewrite this inequality in terms of the geometry of the surface obtained by intersecting the past null cone of the original surface S with a constant time hyperplane and the "time height" function of S over this hyperplane. We also specialize to the case when S lies in the past null cone of a point and show the validity of the corresponding inequality in any dimension (in four dimensions this inequality was proved by Tod). Exploiting properties of convex hypersurfaces in Euclidean space we write down the Penrose inequality in the Minkowski spacetime of arbitrary dimension n+2 as an inequality for two smooth functions on the sphere. We finally obtain a sufficient condition for the validity of the Penrose inequality in the four dimensional Minkowski spacetime and show that this condition is satisfied by a large class of surfaces. △ Less

Submitted 13 March, 2012; originally announced March 2012.

Comments: 25 pages, 2 figures, Latex

Polydispersity Effects in the Dynamics and Stability of Bubbling Flows

Authors: E. Salinas-Rodríguez, R. F. Rodríguez, J. M. Zamora, A. Soria

Abstract: The occurrence of swarms of small bubbles in a variety of industrial systems enhances their performance. However, the effects that size polydispersity may produce on the stability of kinematic waves, the gain factor, mean bubble velocity, kinematic and dynamic wave velocities is, to our knowledge, not yet well established. We found that size polydispersity enhances the stability of a bubble colu… ▽ More The occurrence of swarms of small bubbles in a variety of industrial systems enhances their performance. However, the effects that size polydispersity may produce on the stability of kinematic waves, the gain factor, mean bubble velocity, kinematic and dynamic wave velocities is, to our knowledge, not yet well established. We found that size polydispersity enhances the stability of a bubble column by a factor of about 23% as a function of frequency and for a particular type of bubble column. In this way our model predicts effects that might be verified experimentally but this, however, remain to be assessed. Our results reinforce the point of view advocated in this work in the sense that a description of a bubble column based on the concept of randomness of a bubble cloud and average properties of the fluid motion, may be a useful approach that has not been exploited in engineering systems. △ Less

Submitted 27 September, 2005; originally announced September 2005.

Comments: 11 pages, 2 figures, presented at the 3rd NEXT-SigmaPhi International Conference, 13-18 August, 2005, Kolymbari, Crete