Shobhnik Kriplani

Hi! I am a third-year undergrad pursuing Math & Computing at the Indian Institute of Science, Bangalore. My primary research interests lie in reasoning, alignment, and safety in Language Models.

I'm currently working on AI safety research under the guidance of Prof. Haohan Wang, focusing on improving the robustness and reliability of large language models against adversarial attacks and misuse.

Previously, I worked under the guidance of Prof. Chandan Reddy (VT) and Prof. Amrit Singh Bedi (UCF), where I contributed to SurfaceBench, a large-scale benchmark for evaluating the ability of self-evolving language models to recover symbolic equations of 3D scientific surfaces, and conducted research on inference-time alignment and reward hacking in LLMs, focusing on mathematical analysis of decoding methods like Best-of-N, Controlled Decoding, and related optimization techniques.

My research journey began at the DACAS Lab under Prof. Jishnu Keshavan, where I developed a framework for solving inverse robot kinematics using Invertible Neural Networks (INNs), achieving significant improvement in accuracy.

I'm always excited to explore new ideas. Feel free to send me an email at shobhnik1 [at] gmail.com

News

Arpil, 2026	Gave a lecture on Neural Distance for Biological Sequences. Lecture Slides can be found here.
Feb, 2026	Our paper SurfaceBench has been accepted to TMLR!
Jan, 2026	Exited to start my Research Internship at the University of Illinois Urbana-Champaign!
Dec 8th, 2025	Presented our work on MentorX, a Personalised AI-Tutor, at CASML Conference!
Dec 5th, 2025	Presented our work NanoAgent at Salesforce, Bangalore!
Nov, 2025	My Virginia Tech internship project SurfaceBench: Can Self-Evolving LLMs Find the Equations of 3D Scientific Surfaces? is now on arxiv!
July, 2025	Exited to start my Research Internship at Virginia Tech!
May, 2025	Exited to start my Research Internship at SAFERR AI Lab, University of Central Florida!
Dec, 2024	Started working as a Research Intern at DACAS Lab, IISc Bangalore!
Aug, 2023	Joined the Indian Institute of Science, Bangalore as an Undergrad student in Mathematics and Computing.

Publications

TMLR

SurfaceBench: A Geometry-Aware Benchmark for Symbolic Surface Discovery

Sanchit Kabra, Shobhnik Kriplani, Parshin Shojaee, Chandan K. Reddy.

Abs PDF Code

Equation discovery from data is a central challenge in machine learning for science, which requires the recovery of concise symbolic expressions that govern complex physical and geometric phenomena. Recent large language model (LLM) approaches have shown promise in symbolic regression, yet existing benchmarks predominantly evaluate low-dimensional scalar functions and rely on string-level or regression-based metrics that fail to capture structural and geometric equivalence. We introduce SurfaceBench, the first geometry-aware benchmark for symbolic discovery of three-dimensional surfaces. Unlike scalar curve-fitting tasks, SurfaceBench targets surface-level reasoning, where multi-variable coupling, coordinate transformations, and geometric structure must be inferred directly from data. The benchmark comprises 183 analytically constructed, science-inspired surface equations across 15 categories and three representation paradigms: explicit, implicit, and parametric forms. Each task includes variable semantics and synthetically sampled 3D data, and is designed to stress symbolic composition, structural ambiguity, and representational non-uniqueness while mitigating memorization. To evaluate discovery quality, SurfaceBench incorporates symbolic equivalence checks with geometric metrics of the object-space (Chamfer and Hausdorff distances) and regression-based error measures, allowing the evaluation of functional fidelity beyond algebraic syntax. Empirical evaluation across evolutionary, neural, and LLM-driven frameworks reveals that no current method achieves consistent performance across representation types, with LLM-based approaches exhibiting strong structural priors but limited robustness in parameter calibration and multi-equation reasoning. SurfaceBench provides a challenging and diagnostic testbed that bridges symbolic reasoning and geometric reconstruction, enabling principled benchmarking of compositional generalization and structure-aware scientific induction in high-dimensional equation discovery.