# TRIDENT and SHAKTI: Framework for Hindi AI-Text Detection

> The rapid advancement of Large Language Models (LLMs) has made it increasingly important for organizations to differentiate between human-written and AI-generated content. This differentiation is crit...

**Source**: arxiv.org | **Published**: 2026-11-02 | **Type**: research

## Key Facts

- Differentiate human and AI content to enhance information integrity in your organization.
- Adopt the SHAKTI benchmark to evaluate AI performance in low-resource languages like Hindi.
- Utilize insights from the SHAKTI study to improve language models' human-like text generation.
- Implement targeted prompting strategies to optimize LLM outputs across diverse text genres.
- Invest in multilingual capabilities to stay competitive in global markets and diverse audiences.

## Summary

**Paper:** [TRIDENT and SHAKTI: A Framework and Benchmark for Hindi AI-Text Detection](https://arxiv.org/abs/2610.41232)

**Authors:** Shifali Agrahari, Sujit Kumar, Sanasam Ranbir Singh

\## Executive Summary

The rapid advancement of Large Language Models (LLMs) has made it increasingly important for organizations to differentiate between human-written and AI-generated content. This differentiation is critical for maintaining the integrity of information, especially in multilingual contexts. While much of the existing research focuses on English, there is a significant gap in resources for low-resource languages, such as Hindi. 

To address this need, the researchers developed SHAKTI, a comprehensive benchmark specifically designed for Hindi. This benchmark encompasses four different genres of text and evaluates eight families of LLMs. It employs various prompting strategies and adversarial transformations to assess the ability of these models to generate text that closely resembles human writing. The quality of the SHAKTI benchmark was validated through human annotation, providing insights into the linguistic, grammatical, and sentiment-based differences between texts produced by humans and those generated by AI.

Building on the insights gained from the SHAKTI benchmark, the researchers introduced TRIDENT, a detection framework designed to identify AI-generated content through round-trip translation. TRIDENT comes in three variants: TRIDENT-S, which uses a static threshold; TRIDENT-D, which employs a dynamic threshold; and TRIDENT-C, which utilizes logistic regression informed by multi-distance round-trip translation features. By comparing the semantic consistency of the original text with its translation back from Hindi to English, TRIDENT aims to effectively detect AI-generated content.

The research conducted extensive experiments to evaluate TRIDENT’s performance against conventional and supervised benchmarks. The results indicated that TRIDENT outperformed these baselines, with TRIDENT-S achieving the highest level of accuracy. Additionally, the framework demonstrated its applicability across seven languages and its robustness in various adversarial scenarios, although its effectiveness can vary depending on the translation model used.

This research highlights the importance of developing multilingual benchmarks and detection tools, particularly for languages that have been historically underrepresented in AI research. Organizations could leverage tools like TRIDENT to enhance their content verification processes, ensuring the authenticity of information in a multilingual landscape. This capability may be especially relevant for industries such as media, education, and any sector that relies on accurate communication across different languages.

\## Academic Abstract

As Large Language Models (LLMs) become increasingly multilingual, distinguishing human-written from AI-generated text is essential for maintaining information integrity. However, research remains English-centric, with limited benchmarks for low-resource languages such as Hindi. To address this gap, we introduce SHAKTI, a large-scale Hindi benchmark covering four genres, eight LLM families, multiple prompting strategies, and adversarial transformations. We evaluate its quality through human annotation and analyze linguistic, grammatical, and sentiment-based differences between human- and AI-generated texts. Building on these findings, we propose TRIDENT, a round-trip translation-based detection framework with three variants: TRIDENT-S (static threshold), TRIDENT-D (dynamic threshold), and TRIDENT-C (logistic regression using multi-distance RTT features). TRIDENT leverages semantic consistency between original and round-trip translated texts to identify AI-generated content. Extensive experiments show that TRIDENT outperforms conventional and supervised baselines, with TRIDENT-S achieving the best overall performance. Further evaluations across seven languages, adversarial settings, and unseen LLMs demonstrate its multilingual applicability and robustness, while highlighting sensitivity to the translation model.

## Frequently Asked Questions

**What business problems does this research solve?**

This research addresses the challenge organizations face in differentiating between human-written and AI-generated content, which is crucial for maintaining the integrity and trustworthiness of information in multilingual contexts.

**Which industries benefit most from the findings of this research?**

Industries that rely heavily on content creation and information dissemination, such as media, education, and customer service, could benefit significantly from the ability to accurately identify AI-generated text, particularly in languages like Hindi.

**What are the practical implementation considerations for businesses using this framework?**

Businesses may need to consider integrating the SHAKTI benchmark into their existing content moderation systems and ensuring that their teams are trained to utilize the new tools effectively for evaluating content authenticity.

**What resources or expertise are needed for effective implementation of this research in a business context?**

Organizations may require expertise in natural language processing (NLP) and linguistics to understand and utilize the SHAKTI benchmark effectively, as well as resources for training and maintaining AI detection systems.

**What are the competitive advantages of adopting this framework in business operations?**

By employing the SHAKTI framework, businesses could enhance their credibility and trust with audiences, reduce the risk of misinformation, and improve their overall content quality, potentially leading to a stronger market position.

## Links

- [Read on Welcome.AI](https://welcome.ai/content/trident-and-shakti-framework-for-hindi-ai-text-detection)
- [Original source](https://arxiv.org/abs/2610.41232)

---

Source: Welcome.AI | https://welcome.ai/content/trident-and-shakti-framework-for-hindi-ai-text-detection