Skip to content
Hot eventLive

SyntaxBench:LLM字符级推理统计诊断基准

1 reports1 sources16 hr ago updated

Get the story

AI overview

2026年10月5日,arXiv(Computation and Language)发布一手报道,介绍SyntaxBench——一个面向大语言模型字符级推理的诊断基准与统计评估框架。该框架包含五项核心任务:字符计数、字母包含、回文检测、编辑距离、最长字符串选择;此外还设有更难的子串抽取压力测试index_to_span。目前公开信息仅涉及任务构成与框架定位,暂无模型评测结果或性能数据。

Generated from reports · updated 16 hr ago

Timeline

Follow the coverage from different angles.

Oct 5, 2026
  1. arXiv · Computation and Language
    SyntaxBench:面向大语言模型字符级推理的统计诊断框架

    SyntaxBench 是一个面向字符级推理的诊断基准与统计评估框架,包含字符计数、字母包含、回文检测、编辑距离、最长字符串选择五项核心任务,以及更难的子串抽取压力测试 index_to_span。

Heat trend

Current heat 6·Comparable peak 10(Oct 5)·Comparable change over 24 hours –

02.557.510Oct5Oct5Oct5Oct6

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.