TriEval: A Resource-Efficient Pipeline for LLM Bias, Toxicity, and Truthfulness Assessment
TriEval is a new pipeline designed to assess bias, toxicity, and truthfulness in large language models (LLMs) efficiently. It allows researchers to evaluate multiple parameters simultaneously without requiring extensive computational resources. The tool has been tested on various models and is being released as open source to enhance accessibility for researchers with limited resources.
- ▪TriEval evaluates LLM outputs across multiple parameters including bias, toxicity, and truthfulness.
- ▪The pipeline can run on standard laptops without the need for GPU clusters.
- ▪TriEval has shown clear differences in performance between open-source and closed-source models.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2606.03036 (cs) [Submitted on 2 Jun 2026] Title:TriEval: A Resource-Efficient Pipeline for LLM Bias, Toxicity, and Truthfulness Assessment Authors:Akshatha Srikantha, Manpreet Singh, Yash Jajoo, Shyamal Lakhanpal View a PDF of the paper titled TriEval: A Resource-Efficient Pipeline for LLM Bias, Toxicity, and Truthfulness Assessment, by Akshatha Srikantha and 3 other authors View PDF Abstract:LLMs have evolved from basic chatbots to the backbone of the AI ecosystem, now widely used in healthcare, schools, and government services. The domain-wide adoption of LLMs necessitates continuous evaluation to ensure their safety and fairness.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
