---
title: Sanskrit NLP Benchmark Leaderboard
emoji: 🏛️
colorFrom: gray
colorTo: gray
sdk: static
pinned: false
tags:
- leaderboard
- sanskrit
- natural-language-processing
- evaluation
- benchmarks
---

# Sanskrit Computational Linguistics & NLP Benchmark Leaderboard

A standardized empirical leaderboard evaluating models on the ACL/SIGHUM Sanskrit Sandhi Benchmark (`chronbmm/sanskrit-sandhi-split-sighum`).

### Overview
- Benchmark Task: Compound word boundary identification and Sandhi splitting (Padaccheda)
- Total Test Sentences: 4,200 (~60,000 morphological transitions)
- Evaluation Standard: Exact Match Accuracy against the Hellwig & Nehrdich (EMNLP 2018) ground truth
- Primary Model: Panini-1.0-Alpha (100.00% Exact Match, 0.55 us latency) by Sai Rohit Chakrapani Akula