arXiv subjects

Yony Bresler

Publications and source records attributed to Yony Bresler.

4 recordsLinked to original sources

arXiv2026

Knowledge Distillation with Structured Chain-of-Thought for Text-to-SQL

Deploying accurate Text-to-SQL systems at the enterprise level faces a difficult trilemma involving cost, security and performance. Current solutions force enterprises to choose between expensive, proprietary Large Language Models (LLMs) and low-performing Small Language Models (SLMs). Efforts to improve SLMs often rely on distilling reasoning from large LLMs using unstructured Chain-of-Thought (CoT) traces, a process that remains inherently ambiguous. Instead, we hypothesize that a formal, structured reasoning representation provides a clearer, more reliable teaching signal, as the Text-to-SQL task requires explicit and precise logical steps. To evaluate this hypothesis, we propose Struct-SQL, a novel Knowledge Distillation (KD) framework that trains an SLM to emulate a powerful large LLM. Consequently, we adopt a query execution plan as a formal blueprint to derive this structured reasoning. Our SLM, distilled with structured CoT, achieves an absolute improvement of 8.1% over an unstructured CoT distillation baseline. A detailed error analysis reveals that a key factor in this gain is a marked reduction in syntactic errors. This demonstrates that teaching a model to reason using a structured logical blueprint is beneficial for reliable SQL generation in SLMs.

cs.CL↗

arXiv2022

Improved Batching Strategy For Irregular Time-Series ODE

Irregular time series data are prevalent in the real world and are challenging to model with a simple recurrent neural network (RNN). Hence, a model that combines the use of ordinary differential equations (ODE) and RNN was proposed (ODE-RNN) to model irregular time series with higher accuracy, but it suffers from high computational costs. In this paper, we propose an improvement in the runtime on ODE-RNNs by using a different efficient batching strategy. Our experiments show that the new models reduce the runtime of ODE-RNN significantly ranging from 2 times up to 49 times depending on the irregularity of the data while maintaining comparable accuracy. Hence, our model can scale favorably for modeling larger irregular data sets.

cs.LG↗

arXiv2018

Near the jamming transition of elastic active cells: A sharp-interface approach

We use a sharp interface model for active cells to study the jamming transition point and behavior near it by varying cell concentration, active velocity and elasticity, including a binary mixture of soft and stiff cells. We determine the jamming transition point, as well as behavior near the transition,including the effective diffusion, and sixfold bond correlations. Finally, we expand on previous studies by showing the Voronoi dimensionless cell shape can be treated as an order parameter at any concentration.

cond-mat.soft↗

arXiv2018

Effects of cell elasticity on the migration behavior of a monolayer of motile cells: Sharp Interface Model

In order to study the effect of cell elastic properties on the behavior of assemblies of motile cells, this paper describes an alternative to the cell phase field (CPF) \cite{Palmieri2015} we have previously proposed. The CPF is a multi-scale approach to simulating many cells which tracked individual cells and allowed for large deformations. Though results were largely in agreement with experiment that focus on the migration of a soft cancer cell in a confluent layer of normal cells \cite{Lee2012}, simulations required large computing resources, making more detailed study unfeasible. In this work we derive a sharp interface limit of CPF, including all interactions and parameters. This new model offers over $200$ fold speedup when compared to our original CPF implementation. We demonstrate that this model captures similar behavior and allows us to obtain new results that were previously intractable. We obtain the full velocity distribution for a large range of degrees of confluence, $ρ$, and show regimes where its tail is heavier and lighter than a normal distribution. Furthermore, we fully characterize the velocity distribution with a single parameter, and its dependence on $ρ$ is fully determined. Finally, cell motility is shown to linearly decrease with increasing $ρ$, consistent with previous theoretical results.

cond-mat.soft↗