v1.4.1

中文版

基准测试数据集

• 命名实体识别: 新增 12 个 NER（命名实体识别）数据集
• 语音识别: 新增 TORGO 数据集，用于构音障碍语音识别评测，支持 SemScore 评估
• 多模态评测: 新增 RefCOCO 基准测试
• 代码评测: 新增 Terminal-bench 终端命令能力评测

功能增强

• 性能测试: 新增 SLA 自动调优功能，优化性能测试体验
• 服务模式: 新增异步服务支持和 Gradio UI 界面
• 数据加载: 优化本地 JSONL 数据集加载功能

问题修复

• 修复 HallusionBench 数据加载问题
• 修复流式响应解析中的 SSE 分块处理问题

---

English Version

Benchmark Datasets

• Named Entity Recognition: Added 12 NER (Named Entity Recognition) datasets
• Speech Recognition: Added TORGO dataset for dysarthria speech recognition with SemScore evaluation
• Multimodal Evaluation: Added RefCOCO referring expression comprehension benchmark
• Code Evaluation: Added Terminal-bench for terminal command capability assessment

Feature Enhancements

• Performance Testing: Added SLA auto-tuning functionality to optimize performance testing experience
• Service Mode: Added asynchronous service support and Gradio UI interface
• Data Loading: Optimized local JSONL dataset loading functionality

Bug Fixes

• Fixed HallusionBench data loading issues
• Fixed SemScore computation errors
• Fixed eval_config loading related issues

What's Changed

• [Fix] hallusion_bench load data by @Yunnglin in #1092
• [Feature] Add perf SLA auto tune by @Yunnglin in #1095
• [Feature] add service async and gradio ui by @Yunnglin in #1103
• fix(streaming): Robust parsing of SSE chunks with multiple events and \r\n normalization by @amumu96 in #1102
• Add 12 NER Datasets by @penguinwang96825 in #1106
• [Benchmark] Add TORGO Dataset for Dysarthria Speech Recognition with SemScore Evaluation by @penguinwang96825 in #1107
• [Benchmark] Add RefCOCO by @mushenL in #1109
• [Fix] computation error in SemScore by @penguinwang96825 in #1110
• [Feature] Update load local jsonl by @Yunnglin in #1111
• [Fix] eval_config load by @Yunnglin in #1116
• [Benchmark] Add terminal-bench by @Yunnglin in #1114

New Contributors

• @amumu96 made their first contribution in #1102

Full Changelog: v1.4.0...v1.4.1