arXiv subjects

Rishabh

Publications and source records attributed to Rishabh.

11 recordsLinked to original sources

arXiv2026

Modeling Optical Polarization Evolution in Myelinated Axon Waveguides with Realistic Imperfections

Biophotonic signaling via axons has been proposed as a potential mode of neural communication, where information might be encoded not only in photon number and wavelength but also in polarization. Although earlier computational studies have examined how structural imperfections influence optical transmission, their effects on polarization fidelity remain unexplored; previous modeling of polarization fidelity in myelinated axons has largely focused on idealized geometries. This study incorporates three structural imperfections characteristic of axons in vivo: variation in myelin thickness, non-circular cross-sectional geometry, and axonal bending, within a model that includes four nodes of Ranvier. We find that variation in myelin thickness alone has minimal impact on fidelity, while non-circular cross-sections show strong mode dependence. Axonal bending has the most significant influence, generating large fluctuations and deep fidelity dips. When all imperfections are combined in a single axon model, the simulations show substantial drops in fidelity, yet certain modes exhibit recovery, with repeated revivals reaching values of around 0.8, which exceeds the revivals observed in the single imperfection cases. Overall, the results indicate that although structural imperfections affect polarization, polarization-based biophotonic signals might remain recoverable even in realistic axons, lending support to the plausibility of polarization-based biophotonic signaling in the brain.

physics.bio-ph↗

arXiv2025

Loong: Synthesize Long Chain-of-Thoughts at Scale through Verifiers

Recent advances in Large Language Models (LLMs) have shown that their reasoning capabilities can be significantly improved through Reinforcement Learning with Verifiable Reward (RLVR), particularly in domains like mathematics and programming, where ground-truth correctness can be automatically evaluated. However, extending this success to other reasoning-intensive domains remains challenging due to the scarcity of high-quality, verifiable datasets and the high cost of human supervision. In this work, we introduce the Loong Project: an open-source framework for scalable synthetic data generation and verification across a diverse range of reasoning-intensive domains. The framework consists of two key components: (1) LoongBench, a curated seed dataset containing 8,729 human-vetted examples across 12 domains (e.g., Advanced Mathematics, Chemistry, Logic), each paired with executable code and rich metadata; and (2) LoongEnv, a modular synthetic data generation environment that supports multiple prompting strategies to produce new question-answer-code triples. Together, these components form an agent-environment loop that enables reinforcement learning, where an LLM-based agent is rewarded for generating Chain-of-Thought (CoT) solutions that align with code-executed answers. Empirically, we benchmark LoongBench on a broad suite of both open-source and proprietary LLMs to evaluate domain coverage and reveal performance bottlenecks. In addition, we conduct a comprehensive analysis of synthetic data generated by LoongEnv, examining correctness, difficulty, and diversity. Code and documentation are available at https://github.com/camel-ai/loong.

cs.LG↗

arXiv2025

Distilling Tool Knowledge into Language Models via Back-Translated Traces

Large language models (LLMs) often struggle with mathematical problems that require exact computation or multi-step algebraic reasoning. Tool-integrated reasoning (TIR) offers a promising solution by leveraging external tools such as code interpreters to ensure correctness, but it introduces inference-time dependencies that hinder scalability and deployment. In this work, we propose a new paradigm for distilling tool knowledge into LLMs purely through natural language. We first construct a Solver Agent that solves math problems by interleaving planning, symbolic tool calls, and reflective reasoning. Then, using a back-translation pipeline powered by multiple LLM-based agents, we convert interleaved TIR traces into natural language reasoning traces. A Translator Agent generates explanations for individual tool calls, while a Rephrase Agent merges them into a fluent and globally coherent narrative. Empirically, we show that fine-tuning a small open-source model on these synthesized traces enables it to internalize both tool knowledge and structured reasoning patterns, yielding gains on competition-level math benchmarks without requiring tool access at inference.

cs.LG↗

arXiv2024

Quantum theory of a potential biological magnetic field sensor: radical pair mechanism in flavin adenine dinucleotide biradicals

Recent studies in vitro and in vivo suggest that flavin adenine dinucleotide (FAD) on its own might be able to act as a biological magnetic field sensor. Motivated by these observations, in this study, we develop a detailed quantum theoretical model for the radical pair mechanism (RPM) for the flavin adenine biradical within the FAD molecule. We perform molecular dynamics simulations to determine the distance between the radicals on FAD, which we then feed into a quantum master equation treatment of the RPM. In contrast to previous semi-classical models which are limited to the low-field and high-field cases, our quantum model can predict the full magnetic field dependence of the transient absorption signal. Our model's predictions are consistent with experiments.

physics.bio-ph↗

arXiv2023

Radical pairs and superoxide amplification can explain magnetic field effects on planarian regeneration

Weak magnetic field exposure can affect many biological processes across a wide range of living organisms. Recently, it has been observed that weak magnetic fields can modulate reactive oxygen species (ROS) concentration, affecting regeneration in planaria. These effects show unusual nonlinear dependence on magnetic field strength, including a sign change. In another study by the same group, superoxide is identified as the particular ROS being modulated. We propose a radical pair mechanism based on a flavin-superoxide radical pair to explain the modulation of superoxide production and its effect on planarian regeneration. The results of our calculations favor a triplet-born radical pair. Our yield calculations can reproduce the observed magnetic field dependence, including the sign change. Moreover, to explain the size of the effect on ROS concentration, we suggest a simple amplification model inspired by known biochemical mechanisms and lay out the conditions for such a model to work. Further, we also make empirical predictions concerning the hypomagnetic field effects on planarian regeneration.

physics.bio-ph↗

arXiv2022

Parity-detection-based Mach-Zehnder interferometry with coherent and non-Gaussian squeezed vacuum states as inputs

We theoretically explore the advantages rendered by non-Gaussian operations in phase estimation using a parity-detection-based Mach-Zehnder interferometer, with one input being a coherent state and the other being a non-Gaussian squeezed vacuum state (SVS). We consider a realistic model to perform three different non-Gaussian operations, namely photon subtraction, photon addition, and photon catalysis on a single-mode SVS. We start by deriving the Wigner function of the non-Gaussian SVSs, which is then utilized to derive the expression for the phase sensitivity. The analysis of the phase sensitivity reveals that all three different non-Gaussian operations can enhance the phase sensitivity under suitable choices of parameters. We also consider the probabilistic nature of these non-Gaussian operations, the results of which reveal the single photon addition to be the optimal operation. Further, our analysis also enables us to identify the optimal squeezing of the SVS and the transmissivity of the beam splitter involved in the implementation of the non-Gaussian operations.

quant-ph↗

arXiv2022

Enhanced phase estimation in parity detection based Mach-Zehnder interferometer using non-Gaussian two-mode squeezed thermal input state

While the quantum metrological advantages of performing non-Gaussian operations on two-mode squeezed vacuum (TMSV) states have been extensively explored, similar studies in the context of two-mode squeezed thermal (TMST) states are severely lacking. In this paper, we explore the potential advantages of performing non-Gaussian operations on TMST state for phase estimation using parity detection based Mach-Zehnder interferometry. To this end, we consider the realistic model of photon subtraction, addition, and catalysis. We first provide a derivation of the unified Wigner function of the photon subtracted, photon added and photon catalyzed TMST state, which to the best of our knowledge is not available in the existing literature. This Wigner function is then used to obtain the expression for the phase sensitivity. Our results show that performing non-Gaussian operations on TMST states can enhance the phase sensitivity for significant ranges of squeezing and transmissivity parameters. We also observe that incremental advantage provided by performing these non-Gaussian operations on the TMST state is considerably higher than that of performing these operations on the TMSV state. Because of the probabilistic nature of these operations, it is of utmost importance to take their success probability into account. We identify the photon catalysis operation performed using a high transmissivity beam splitter as the optimal non-Gaussian operation when the success probability is taken into account. This is in contrast to the TMSV case, where we observe photon addition to be the most optimal. These results will be of high relevance for any future phase estimation experiments involving TMST states. Further, the derived Wigner function of the non-Gaussian TMST states will be useful for state characterization and its application in various quantum information protocols.

quant-ph↗

arXiv2022

Realistic non-Gaussian operations scheme in parity detection based Mach-Zehnder quantum interferometry

We theoretically analyze phase sensitivity using parity detection based Mach Zehnder interferometer (MZI) with the input states generated by performing non-Gaussian operations, viz., photon subtraction, photon addition, and photon catalysis on a two-mode squeezed vacuum (TMSV) state. Since these non-Gaussian operations are probabilistic, it is of utmost importance to take the success probability into account. To this end, we consider the realistic model of photon subtraction, addition, and catalysis and derive a single expression of the Wigner function for photon subtracted, added, and catalyzed TMSV state. The Wigner function is used to evaluate the lower bound on the phase sensitivity via quantum Cramer-Rao bound and parity detection based phase sensitivity in MZI. We identify the ranges of squeezing and transmissivity parameters where the non-Gaussian states provide better phase sensitivity than the TMSV state. On qualitatively taking the success probability into account, it turns out that the photon addition is the most advantageous non-Gaussian operation. We hope that the generalized Wigner function derived in this work will be useful in various quantum information protocols and state characterization.

quant-ph↗

arXiv2022

Evolution of two-mode quantum states under a dissipative environment: which quantum resource survives better, squeezing or entanglement?

We explore the relative robustness of squeezing and entanglement (which are quantum resources interconvertible via passive optics) for two-mode Gaussian states under different dissipative environments. When the individual modes interact with identical local baths, entanglement and squeezing decay at the same rate. However, when only one of the modes interacts with a local bath, the comparative robustness of entanglement and squeezing depends on the initial squeezing of the state. Similarly, when the system interacts with a global bath, the robustness of entanglement and squeezing depends on the initial squeezing. Thus depending on the nature of dissipative environments and the initial squeezing of the state, one can select the more robust form of resource out of squeezing and entanglement to store quantumness. This can be used to effectively enhance the performance of various quantum information processing protocols based on continuous variable Gaussian states.

quant-ph↗

arXiv2021

Radical pairs may explain reactive oxygen species-mediated effects of hypomagnetic field on neurogenesis

Exposures to a hypomagnetic field can affect biological processes. Recently, it has been observed that hypomagnetic field exposure can adversely affect adult hippocampal neurogenesis and hippocampus-dependent cognition in mice. In the same study, the role of reactive oxygen species (ROS) in hypomagnetic field effects has been demonstrated. However, the mechanistic reasons behind this effect are not clear. This study proposes a radical pair mechanism based on a flavin-superoxide radical pair to explain the modulation of ROS production and the attenuation of adult hippocampal neurogenesis in a hypomagnetic field. The results of our calculations favor a singlet-born radical pair over a triplet-born radical pair. Our model predicts hypomagnetic field effects on the triplet/singlet yield of comparable strength as the effects observed in experimental studies on adult hippocampal neurogenesis. Our predictions are also in qualitative agreement with experimental results on superoxide concentration and other observed ROS effects. We also predict the effects of applied magnetic fields and oxygen isotopic substitution on adult hippocampal neurogenesis. Our findings strengthen the idea that nature might harness quantum resources in the context of the brain.

q-bio.NC↗

arXiv2021

CGuard: Efficient Spatial Safety for C

Spatial safety violations are the root cause of many security attacks and unexpected behavior of applications. Existing techniques to enforce spatial safety work broadly at either object or pointer granularity. Object-based approaches tend to incur high CPU overheads, whereas pointer-based approaches incur both high CPU and memory overheads. SGXBounds, an object-based approach, is so far the most efficient technique that provides complete out-of-bounds protection for objects. However, a major drawback of this approach is that it can't support address space larger than 32-bit. In this paper, we present CGuard, a tool that provides object-bounds protection for C applications with comparable overheads to SGXBounds without restricting the application address space. CGuard stores the bounds information just before the base address of an object and encodes the relative offset of the base address in the spare bits of the virtual address available in x86_64 architecture. For an object that can't fit in the spare bits, CGuard uses a custom memory layout that enables it to find the base address of the object in just one memory access. Our study revealed spatial safety violations in the gcc and x264 benchmarks from the SPEC CPU2017 benchmark suite and the string_match benchmark from the Phoenix benchmark suite. The execution time overheads for the SPEC CPU2017 and Phoenix benchmark suites were 42% and 26% respectively, whereas the reduction in the throughput for the Apache webserver when the CPUs were fully saturated was 30%. These results indicate that CGuard can be highly effective while maintaining a reasonable degree of efficiency.

cs.CR↗