We hope L-Eval could help researchers and developers track the progress of long-context language models (LCLMs) and understand the strengths/shortcomings of different methods. …

Forgot Password?

Leval provides the open source tools and frameworks your team needs to bootstrap SaaS products lightning fast. Skip the setup, focus on building what matters. © 2025 Leval. All rights …

Designed for use with compatible Lafayette Instrument Company Evaluation systems, such as the Jackson Strength Evaluation System, Digital Hand Dynamometer, LEval provides a wireless …

We start with Leval-S, a benchmark focused on measuring gender bias and stereotyping in model outputs. Additional benchmarks are under development to expand our scope.

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

Jul 20, 2023 · To bridge this gap, we propose L-Eval to institute a more standardized evaluation for long context language models (LCLMs) addressing two key aspects: dataset construction …

Jan 8, 2026 · The Leval-S benchmark measures the degree of gender biases present in model outputs. It helps assess how fairly models treat different genders in their generated text and …

The input file denoted by --path should be a jsonl file like the examples in LEval-data folder. In this case, we annotate two new QA pairs based on the long input. After clicking submit, the results …

In this repo, we implement the evaluation of Ada-LEval on GPT-4-Turbo-0125 (an example for APIs) and internlm2- [7b/20b] (an example for opensource LLMs). You can follow our …